Current fontforge changes
Older fontforge changes

• 1-Mar-2004
  • The Check directions option of Find Problems caused multilayer mode to crash.
  • Expand stroke would produce counter-clockwise paths when given an open contour and asked to fill it with a circular (or elliptical) pen
  • Try to do a better job of guessing the correct direction of a path in an eps file (or a type3 or svg font)
  • FontForge would crash when editing in the grid layer
  • When the control key is down, make the scroll wheel do a magnify/minify.
  • Various cosmetic changes to find problems
    1. Specify that "nearness" is measured in em-units
    2. change "refs deeper than" to "refs nested deeper than"
    3. make check dir be on by default
  • Put in more comments about fontforge. We now have a domain under fontforge, future releases will be found there.
• 29-Feb-2004
  • FontForge could crash if given a bad 'cmap' subtable
  • Clean up generation of OS/2 a bit
  • Pass motion events to windows even when a dlg is active (so that the mouse location will be shown in an outline window when the Point info dlg is active).
  • Put in a comment about changing to FontForge in the copyright notice the program prints at start up.
  • The "preflight" script on the mac had problems if the program had never been installed before
• 27-Feb-2004
  • Redid the way menus respond to arrow keys
  • a multilayer enabled FontForge would crash when editing a normal font.
• 24-Feb-2004
  • Remove the AddHint command and replace it with AddHHint and AddVHint.
  • FontForge would crash when generating a mac family with NFNT bitmaps.
  • Added support of exporting a character into PDF format.
  • Oh dear, I was figuring out hintmasks in the wrong direction (because PS splines get reversed before output, so I now reverse them during the hintmask calculation too).
  • At Werner's suggestion, added code to show what hints are currently active at a point.
  • At Werner's suggestion added code to show what points are affected by a hintmask
  • Added support for real editing of type3 (and svg) fonts. Support for editing stroked paths, different fill colors, multiple stroke/fill operations, etc. Because this involves a lot of extra overhead it is wrong for people who want to edit CJK fonts, so this mode is off by default. It may be turned on by

    $ configure --with-multilayer
    

  • FontForge failed to interpret h/vstem3 commands properly when building a hintmask (when reading a type1 font with hint substitution)
• 11-Feb-2004
  • FontForge would crash when saving a non-multi master font.
  • Adding new scripting commands ClearCharCounterMasks, SetCharCounterMask
• 10-Feb-2004
  • FontForge could produce bad tfm files.
  • Added CheckForAnchorClass scripting command
  • Added SetFontHasVerticalMetrics scripting command.
  • FontForge would crash if asked to generate a contextual lookup which refered to a lookup which only was activated by a character(s) which was not worth outputting.
  • FontForge did the wrong thing when evaluating the PostScript "index" command (it returned the wrong value from the stack (stack index off by one))
  • Added support for the PostScript "for" loop.
  • I have created a simple (1 axis) Multiple Master Type1 font with FontForge. It works with both ghostview and freetype.
  • FontForge would crash when generating a ttf file from an sfd file which contain an invalid script/lang index. Put in code to warn user about this, and in some cases to correct it.
  • FontForge had problems reading sfd files in the old point count format, this meant we still got bad point count errors.
  • FontForge wrote bad sfd files if there were countermasks (all the contours of the glyph would be lost)
  • I finally figured out why FontForge kept crashing when palettes were docked and an outline view was closed.
  • Fixed various crashes in Interpolate fonts
  • FontForge had problems with using subroutines to express references in PostScript (introduced 30-Jan)
  • Added a new scripting commands AddHint and ReplaceCharCounterMasks.
• 4-Feb-2004
  • Make it illegal to select both "Condense" and "Extend" in the mac style.
• 3-Feb-2004
  • Point Info dlg crashed on the mac
  • Add UI control over Mac Style through the Font Info dlg
  • more prep work for mm fonts
• 2-Feb-2004
  • Generating a postscript font from a database with quadratic splines didn't work
  • FontForge crashed on a font with two /CharStrings entries
  • Small tweak to menubar behavior given left/right arrow (for Werner)
  • Composite glyphs created with seac (type1) had the wrong transformation for the accent.
  • more prep work for mm fonts
  • Still have problems with ttf point counts
  • If a user loaded a ttf file and made it monospace and then generated it as ttf, all glyphs except the .notdef had an advance width of 0
• 1-Feb-2004
  • sfd reader had a bug. Went into an infinite loop if a hintmask contained any of the characters [a-fA-F].
• 31-Jan-2004
  • In Type2 output, when putting a glyph into a subroutine, if that glyph had no conflicts but had a hintmask fontforge produced a bad font (which crashed FontForge when it tried to read it in).
  • configure script still wasn't working on solaris (or mac with libiconv probably)
• 30-Jan-2004
  • The kerning state machine addition broke the indic state machine dialog
  • The mac docs claim that the 'feat' table should be sorted by feature id. But if you do this WorldText can't find features. All Apple fonts seem to be sorted by the order the features are used in the 'morx' table's chains. WorldText works if you do this.
  • Yannis proposed several extensions to the State Machine editor, most notably the ability to move the edit dialog with Up/Down/Left/Right buttons.
  • I had the feature settings for Linguistic Rearrangement and Vertical Subs. reversed.
  • In some cases I wasn't outputting anything to turn a feature off.
  • Fixed many problems with kerning state machine dlg
  • Fixed several problems with insertion state machines
  • Fixed several problems with reading in state machines.
  • Added SetMacStyle() to give the user control over how mac families are built.
  • Added a fflush after doing Print()s in scripting
  • FontForge would crash when given an extremely long real number in a cff dictionary (in this case the font itself was in error)
  • FontForge did not copy a bitmap character's width when it copied the bitmap.
  • Char Info didn't handle pairwise positioning properly.
  • Some more pangrams from http://shair.net/misc/txt/pangram.en (link provided by David Shaal)
  • My old approach to hint substitution caused problems with multiple master fonts. So now I keep track of which points are substitution points and what the hint masks are at those points. There is a new command which will automatically figure out these points, and the user may control them directly through Point Info (Element->Get Info). At the same time I figured I might as well do the same thing for counter masks, so FontForge will read counter masks in from otf file (I did not implement this for type1s), there is a new command to try and make a guess as to what they should be, and the user may control them directly through Element->Char Info. (new scripting commands, SubstitutionPoints() and AutoCounter() do the obvious).
  • Added a Hints->Don't AutoHint command to mark a character as having hints the user likes which should not be modified until the user explicitly calls AutoHint(). (and a matching scripting command).
  • Added ability to control whether a PostScript font (type1&type2) is output with rounding.
  • FontForge could generate a bad 'post' table for bitmap only fonts.
  • FontForge didn't allow you to create two pairwise positioning items with the same tag.
  • Hungarian names for mac features and for some standard names like "Bold", etc. provided by Laszlo Karoly.
  • Pierre HANSER noticed that if FontForge mapped a single glyph to two encodings then bitmap generation through freetype got confused.
  • If a font contained a non-.notdef character at encoding 0, then generating an opentype font from it had a fair chance of crashing.
  • Added a couple of new select menu items: Select->Contours, Select->First Point, Next Contour
  • Added a new sub-menu to allow ordering contours within a glyph. This is almost totally useless (the final look of the character should not be affected by the order in which the contours are drawn), but for multiple master fonts, the contour list must have the same ordering in corresponding glyphs.
• 11-Jan-2003
  • A patch for reading broken mac tables got lost. I've reinstated it.
• 10-Jan-2003
  • FontForge would crash when asked to use the truetype debugger on fpgm
  • FontForge always coloured the background of debugger windows grey.
  • The Points window did not show any points.
  • Changed the Points window and the character view to show whether a point has a watch point set on it.
  • Display the code range ('fpgm'/'prep'/Glyph) in the character's info line when debugging.
  • When creating a 'fpgm'/'prep' table we created one with some garbage instructions in it.
  • Somehow I neglected to put in a dlg for editing the 'cvt ' table. This has been rectified.
  • There was a bug in the postscript interpreter (the thing used for reading .eps files and type3 fonts) and transforming a point by a matrix with diagonal entries (rotation, skew, etc.) gave the wrong value.
  • Added code to preserve more of the PS graphic state when eps file does a gsave
  • Merge Fonts would crash when merging a font with references into an empty font
  • I wasn't providing default values for width/weight classes in the OS/2 table
  • The bitmap dlg could write one byte beyond the area allocated in some cases.
  • Changed the outline character popup menu to include an Add Anchor Point entry
  • The Ligature index wasn't correctly enabled in some cases.
  • Add little arrows near the first point of each contour (in the outline character view) showing the direction of the contour.
  • The mac's interpretation of the CFF spec is very picky, but correct. If a font contained no references then FontForge would generate an empty subroutine index. This is wrong. There should be no subroutine index, rather than an empty one. (Strangely, global subroutines have the reverse requirement. Stupid). Now if there are no references, then FontForge will not generate a local subroutine index.
  • Yannis Haralambous has corrected the French UI.
  • Three new scripting commands: AddAnchorClass, AddAnchorPoint, RemoveAnchorClass
  • Show ATT did not show glyph classes properly.
  • Made scripting dlg be resizable.
  • Save As & Import dlgs looked bad in French. Various other improvements to make things look better on non-english systems.
  • Added the ability to set the glyph class from Char Info.
  • FontsInFile did not work on pfb files (nor on cff)
  • Opening a mac resource file with multiple FONDs from a script would crash.
  • In Font Info->TTF Names the languages now appear in unicode ordering no matter what the locale (used to be the langs were in alphabetical ordering for English, even if they were actually French)
• 2-Jan-2004
  • More accurate round joints in stroking.
  • Better approximations to splines which turn through 180 degrees or more
  • Support for reading cff files.
  • Yet more problems counting ttf points
  • Update copyright notices.
  • Add code to protect FontForge against reading bad Apple lookup tables (I think the Zapfino.dfont that ships with 10.3 has two bad lookups, but perhaps I just don't understand them).
  • Try harder to avoid using 'DFLT' script when coming up with a default script for glyphs with no innate script.
  • Add a Find Problems to detect 'DFLT'
  • I've been generating bad Simple Positioning subtables for GPOS all along.
  • Be pickier about what glyph names I let the user choose.
  • On cygwin (perhaps elsewhere, but I was never able to find the bug elsewhere) if you invoked one command from the menu while another was still executing, then fontforge would probably crash.
  • Support for reading and writing bare cff files.
  • Build Accents crashed when asked to build accented chars for iacute, etc. in a font without dotlessi.
• 28-Dec-2003
  • RemoveATT scripting command didn't work if the third argument were "*"
  • FontForge's clipping in the character view left something to be desired.
  • Make Meta-Arrow move selected points by 10 times what Arrow would move them by.
  • Grabbed the feature ordering for GPOS/GSUB specified on the microsoft site and taught it to FontForge.
  • Change the Point Info dlg
    • to allow the user to control the point type from there.
    • to enforce constraints on control points (ie. the control points of a curve point must be in line with the point itself)
    • by adding an alternated mode of entering data, by distance and angle (rather than just offset)
  • Some MF fonts have glyph names with characters which aren't allowed in PostScript names
  • FontForge failed to set widthset when it read in glyphs from a windows FNT (FON) file. This meant that some glyphs might be ignored.
  • The context chaining dialog had a bad test for valid glyph names and complained about any name containing digits
  • The histogram dlg would crash if the mouse moved across it when the window was so wide that it could contain more stem widths than were actually in use.
  • Added three commands to the metrics view view menu to allow users to insert glyphs by name or encoding.
  • Yannis points out that afm files produced on fonts with vertical metrics did not contain "WY" values (vertical advance width), also that the vkerns in the file actually contained the values of horizontal kerns.
  • The 'kern' table did not contain pairwise vertical kerns (it did contain vertical kerning by class).
  • The View->Substitutions sub menu of the metricsview was not being cleared properly if the current glyph had no substitutions.
  • Yannis wants the View->Substitutions menu to have a way to return to the original glyph.
  • Yannis points out that the metricsview in vertical mode changed the top side bearing in an unexpected way. Also the vertical advance width did not refresh the display properly.
  • Thanks to David Opstad, I now have an example of a format 1 kerning table and have extended FontForge to support kerning by state machine.
  • The code for generating and loading otb (OpenType bitmap) fonts was broken.
  • And the code for generating my own weird bitmap only fonts for windows had some flaws too (different flaws).
  • If I typed a character into a fontview (or char view) and the font did not contain the character, fontforge would not select that slot that encoded the character, even if the character was in the encoding.
  • When loading a ps type3 font, make sure that the glyph name is available when parsing glyphs so we can produce reasonable error msgs. (mostly this means "Spline stroke is too big"
  • Only generate Spline stroke too big errors once per font.
  • Adam Goode tells me that the format 12 (32bit unicode) cmap was wrong, and provided a patch to fix it.
  • Pretty much rewrote the stroking command from scratch.
  • Merge fonts would crash when merging in a font containing either a ligature caret or a nested lookup.
  • Made FontForge cognizant of the proposed unicode dotlessj character (0237), it will now use either this or the dotlessj in adobe's private use area (depending on which is in the font)
  • FontForge would crash when trying to build an accented character based on inverted breve in pala.ttf
    • It wrongly considered pala.ttf to be slightly italic (.0005°)
    • It would crash when doing this in an italic font

    ---

  • When importing postscript (type3 fonts, eps files) counter-clockwise paths which are both stroked and filled cause problems
  • Changed the version stamp of otb fonts from 'OTTO' to 1.0
• 10-Dec-2003
  • When there were multiple features with the same tag (different scripts with 'kern' for instance) FontForge would produce multiple copies of the same lookup instead of different lookups.
  • FontForge was not setting the right to left lookup flag bit properly for hebrew (and presumably arabic) kerning
  • Remove overlap was coalescing some splines that it should not have.
  • Typo in expand stroke meant that some round line joins were really weird
• 5-Dec-2003
  • When changing the ascent/descent with fontinfo or scaletoem some of the OS/2 fields did not get updated properly.
  • The OS/2 TypoDescent field was set positive instead of negative.
  • Added two scripting functions SetTTFName and GetTTFName to give scripts access to the truetype names as well as the PS ones
  • Patch from Kazuyuki. SetPanose didn't work when argument was an array.
  • New French translations from Pierre
  • $copyright didn't return the correct thing (Patch from Giuseppe Ghibo)
  • Kerning pair data for non-ASCII unicode characters did not get output properly into SVG fonts
  • The AddATT scripting command did not add ligatures correctly leading to problems when time came to generate a font.
• 23-Nov-2003
  • FontForge would crash if given an encoding for the locale which it didn't understand
  • Center/Thirds in width from the font view would change the width
  • There was a synchronization problem and the truetype debugger window could deadlock
  • FontForge failed to read a Contextual (but not Contextual chaining) Coverage lookup properly.
• 10-Nov-2003
  • FontForge was crashing when loading a bitmap font with encoded characters outside the range specified by the encoding (ie. a character 0x10000 in a unicode BMP font).
  • Fixes for several memory leaks provided by Kazuki Ohta
  • Slight improvements to the View->Combinations->Ligature dlg
• 8-Nov-2003
  • Bug introduced 6-Nov, FontForge generated a bad 'loca' table when outputting a ttf outline font with bitmaps.
  • FontForge was getting the truetype point number wrong when reading outlines from an sfd file.
  • If the loca table was 2 or 4 bytes long, FontForge failed to warn about a bad glyph count.
  • FontForge got confused if two truetype glyphs had the same name.
  • When dumping out embedded ttf bitmaps, FontForge got confused by the extra glyphs (.null and return) and generated garbage.
  • Make View->Combinations->ligature dlg a bit better for non-latin glyphs
• 6-Nov-2003
  • The clut used for drawing anti-alias bitmaps in the metrics view was wrong.
  • In a bitmap only ttf file, FontForge was usually producing bad line spacing data as far as the mac was concerned.
  • When importing bitmaps into an empty font, we might as well change to font view to show the bitmaps rather than the empty outline data.
  • If glyph 0 was selected in the fontview then the File->Open Outline command was disabled.
  • With a user defined encoding we were again getting occasional crashes. (introduced by my work on bitmaps)
  • On certain glyphs, the optimizer introduced an slight error when calculating a spline's coefficients from its control point. Even though this error was on the order of 7e-8 it was enough that the rasterizer produced grossly incorrect results. I have pulled the relevant routine out into its own file which is now compiled by itself without optimization.
• 4-Nov-2003
  • When FontForge removes glyphs with Font Info->Encoding->Number of characters, it will now run through the font and remove any ligatures, substitutions, etc. refering to non-existent glyph names.
  • Some improvements to remove overlap.
  • Added a warning message so that when user attempts to output a ttf file where no characters have unicode encodings, then the user at least gets warned that the result will probably not work. For fonts with fewer than 256 glyphs it will offer the choice of a symbol encoding.
  • If a reference were copied in the outline character view and pasted into the font view, fontforge would crash.
  • the expand stroke patch on 29-Sept broke calligraphic and ellipse modes.
  • The mouse's scroll wheel did not work in the metrics view.
  • The metrics view resized very slowly under window managers which do continuous resizing.
  • Give the magnifying glass's rectangle mode a threshold so that it doesn't resize tiny rectangles which are really caused by mouse jitter.
  • Attempt to set a zoom size for character and bitmap windows, but modern window managers don't seem to support the old protocol, and I can't find docs on a new version.
  • Was not converting filenames to unicode properly for the open/generate progress dlgs.
  • FontForge would crash when attempting to draw a character for which in had no (X11) glyph when doing bidirectional text.
• 29-Oct-2003
  • When removing an ATT feature from a glyph using a script, FontForge would screw up memory. Leading to unpredictable results.
• 28-Oct-2003
  • FontForge would crash when converting from cubic to quadratic splines in a script
  • FontForge would leave a dangling pointer when converting from cubic to quadratic. This did not affect ttf generation, but would affect most other operations.
• 27-Oct-2003
  • FontForge would crash when changing the encoding of a font with a .notdef character to a user defined encoding.
  • If one had multiple ligatures for the same glyph with the same tag and script, FontForge would eat all but one of them.
  • When generating most kinds of font, of a font which had ligatures containing glyph names that were not in the font, then FontForge would often screw up memory and eventually crash.
• 26-Oct-2003
  • Simplify sometimes removed points at extrema even when it should not have
  • If one had a ttf font (with instructions) and merged in a postscript font and then generated a ttf font, FontForge would complain about an internal error, but go on to produce a valid font. It should no longer complain.
  • When generating fonts from a script, if the font were in compacted format the generated font would have a compact encoding (instead of the desired encoding).
  • Kanou provided a patch to speed up importing bitmap fonts.
• 23-Oct-2003
  • The change on the 20th to make custom encoding support 32bit unicode was incomplete and made custom encodings crash instead.
  • The File->Display dialog ignored the left-sidebearing of a glyph causing output of some fonts to look awful.
  • Was getting the AdobeStandardEncoding a bit wrong (after the 5-Oct change to the new unicode glyph list which does not contain "fi" nor "fl")
  • I used to set the "widthset" bit on .notdef characters, but there's really no point to that.
  • Maxim Iorsh tells me that U+fb4f should be a historic ligature.
  • On some displays right scrolling the Histogram dlg would leave black lines on the window.
  • AutoHint removed too many stems from consideration as hints on some very bold fonts
  • Make FontBBox of Type1 fonts be executable. Adobe says it doesn't have to be, but dvips expects it to be.
• 20-Oct-2003
  • Windows XP & 2000 insist on having (what cygwin describes as) the execute bit of the file set. So FontForge now sets execute permissions on ttf, otf and pfb fonts on the cygwin platform.
  • If a menu was so big that it had to be positioned over the menubar (instead of above or below it), and the user clicked on the menubar to invoke it, then the mouse release would go to the new menu and invoke whatever item happened to be under the cursor. Not good.
  • Oops. The Effects sub-menu should have been disabled for quadratic (truetype) fonts.
  • Added a new configuration option --with-regular-link to disable my dlopening of libraries. I dislike this. I don't think it is anywhere near as flexible as the dlopen method. It makes it much harder to ship something that will work with freetype (detecting whether the bytecode interpreter is present must be done at link time rather than run time and that is a severe disadvantage. But I have been asked for it.
  • Change internal format so that custom encodings can handle 32 bit unicode.
• 16-Oct-2003
  • FontForge was still outputting a disordered name table.
  • The source tarball on the website included many files twice and some files that it should not have included at all. I should now be less than half the size it used to be.
• 15-Oct-2003
  • FontForge did not bother to open files in binary mode. If cygwin was configured with DOS linebreaks, then fontforge failed to generate valid font files on cygwin.

    ---

  • Added some more warnings to the default build list and fixed the problems they indicated.
• 12-Oct-2003
  • Several bad bugs in last night's build (name table was (usually) all screwed up, conversion from PS to TTF got incorrect Internal Error msgs)
  • Order tables in truetype output

    ---

  • My 5-Oct-2003 libpng fix broke it completely. Should now be refixed.
• 11-Oct-2003
  • Another patch from Kanou to recognize encodings better in bdf files.
  • Various memory fixes found by valgrind
  • The Element->Find Problems->ATT->Missing glyph names fixup dlg had some formatting problems and failed to skip properly.
  • Draw an "x" through encodings which do not map to glyphs
  • If the user is in a compacted encoding and asks to print a font, then revert to the uncompacted encoding for the duration of the printing.
  • View->Show ATT would show ligatures (and substitutions) even if some of the glyphs in the glyph list were missing.
  • Fix a number of bugs found by the MS Font Validator
    • I misunderstood (and the spec incorrectly documented) the behavior of the maxp table
    • I had a typo in figuring out the OS/2 table
  • I had one more case where I was computing the point count of a glyph incorrectly.
  • Make the width line snap to zero when the user drags near it.
• 6-Oct-2003
  • Give the Metrics View a vertical mode in which it does vertical kerning, etc.
  • Add a hack to allow FontForge to parse a weird pfb file.
• 6-Oct-2003
  • Fixed a number of bugs in the new kerning class dlg including:
    • Vertical kerning classes were displayed horizontally
    • When adding a new class the new offsets were garbage
  • The routine to convert a horizontal kerning class to the equivalent vertical kerning class (of the vertical variants of the glyphs in the horizontal class) didn't work.
  • In Back and Lookup Classes of a contextual chaining substitution by class, the [New] button was never enabled. (it was disabled if []SameAs was on, and then never enabled).
  • My new 94x94 encodings were actually 93x93. Patch by kanou.
• 5-Oct-2003
  • Got rid of GreekFixup and added Interpretations of encodings to FontInfo
  • FontForge failed to output ligature substitutions nested under a context/chain sub
  • If a font had more than about 5 nested ligatures, fontforge would screw up its memory and (likely) crash
  • Fixed various problems pertaining to adding a kerning class to an empty font
  • Add a runtime check for the case where the user's system has a dynamic libpng installed without libz
  • If a character had ligature carets then FontForge could crash when determining its script
  • FontForge was reading mac 'kern' format 3 incorrectly.
  • Rewrote the Kerning class creation dialog. I hope it is clearer now.
• 2-Oct-2003
  • got rid of my strange 94x94 encodings for JIS/KSC/GB (based on the kuten array) and used instead the much more obvious encoding that everyone else uses.
  • BDF files were output with the wrong encoding name for JISX0208 (I was outputting "JISX208" instead-- missing a "0").
  • Autotrace failed to read potrace output properly if there were more than one contour in a given glyph.
  • Upgraded psunicodenames to the current Adobe Glyph List for New Fonts.
  • Simplify now works for quadratic splines. If one uses the []Allow Slopes to Change checkbox then results are not good. I need a better algorithm.
  • If asked to generate a type0 postscript font for a font encoded with an extended 1byte encoding (ie. one that had about 300 characters) FontForge would probably crash.
• 29-Sept-2003
  • When a path was both stroked and filled (in a postscript/eps file), fontforge would ignore the stroke.
  • I forgot that the stroke linewidth needs to be transformed by the current transformation matrix. This meant that when a path should have been stroked by an ellipse (when reading a ps/eps file) fontforge would blithely stroke it with a circle.
  • Added some support for reading the /Metrics array.
  • If there were a sharp corner (where the "radius of curvature" of the corner was less than the radius of the pen used to stroke the path) then FontForge would generate an incorrect path.
  • Round line joins were not a good approximation to circular arcs.
  • When filling the outline character view, rounding errors caused the fill bitmap to stray from the outline.
  • Extend the functionality of the magnifying (or minifying) tool so that if the user drags out a rectangle with it then the view will be shifted and scaled so that rectangle fits in the window.
  • FontForge will no longer auto-generate hints on non-integral boundaries. In Type1 fonts such hints become meaningless and just confuse things.
  • FontForge will now draw a halo around points which will be part of a flex hint. See the UpdateFlex preference item.
  • FontForge now has as set of commands for manipulating typographic features in large lumps rather than one at a time.
    • Copy features from one font to another
    • Remove certain features
    • Give features a new tag
  • I have greatly rewritten the code for dealing with AAT fonts
    • FontForge now supports mac feature/setting values to be used as feature tags as well as opentype 4 letter tags
    • FontForge will now read any feature in even if it doesn't map to an opentype tag
    • FontForge understands the most of the state machines in the morx/mort tables (it does not understand ligature state machines, it parses these for all unconditional ligatures, but it will miss any conditional ones).
    • FontForge now lets you edit apple state machines
    • FontForge lets you specify feature/setting names in the font to override the default values specified in preferences
  • Added support for reading Apple's format 3 kern sub-table format (another way of storing a kerning class)
  • Oops. SVG didn't read relative movetos properly in a path option. Patch by Kengo Ichiki.
  • Simplify had a bug when passed a path which consisted of a single point with a zero length spline connecting it to itself
  • Simplify was not greyed out properly in quadratic fonts
  • Some of the low-level error routines consumed massive amounts of cpu time while waiting for events
  • The PostScript parsing routines would sometimes think that "noaccess" was part of the value of something in the private dictionary.
  • Several other patches to importing images by templates from scripts. By Kengo Ichiki
  • Importing pcf bitmaps into the background had some problems.
• 4-Sept-2003
  • the 17-Aug addition of context/chaining broke anchor classes. Should be fixed now.
  • Károly László has added Hungarian translations for the style codes to Font Info->TTF Names
  • If a font had comments, an svg file generated from it had syntax errors.
  • Add support for generating multilingual strings (in 'name') for the apple platform.
  • Add support for the packed gb2312 which mac and windows seem to use.
  • Make a UniqueID of -1 in Font Info mean that the font has no UniqueID.
• 31-Aug-2003
  • Add problems for maximum number of hints (96) and maximum number of subroutine nests (references) ~10.
  • I looked at AppleChancery again and figured out why I was having trouble parsing its 'mort' ligature tables, and put in code to make it easy to parse tables of this type.
  • After figuring that out, I realized there was a bug in my own table generation. Suppose we have two ligatures "f l => fl" and "s t => st", and are given "fst". Then my table would fail to find the "st" ligature. The state of 'read "f" waiting for "l"' would see that it didn't get an "l" and revert back to the initial state. It should instead have checked to see if "s" could start a ligature in its own right, and if so jump to the 'got "s" waiting for "t"' state.
  • The same fix needs to be done for contextual glyph substitution.
• 29-Aug-2003
  • FontForge failed to display the morx table if all it contained were simple substitutions.
  • FontForge generated incorrect simple substitution (AAT) tables if it found a substitution with no valid elements
  • New problem for find problems. Check for characters with more than 1500 points in them.
  • FontForge will now draw selected points in a different color (currently yellow), and draws them bolder. This makes them stand out better for Find Problems (but also makes it harder to distinguish between corner and curved points).
  • Redid the way FontForge handled arabic forms in AAT. Old method was cleaner but depended on behavior which is unspecified in Apple's docs.
• 27-Aug-2003
  • The preference dlg has been broken since about 12 July. Anything on the Preference->Font Info tab is probably wrong.
  • FontForge wasn't always getting the script/lang right when filling in the fields of the New Substitution dlg.
  • The newly generated apple state machines for context/chaining were quite wrong. I hope I've got them right now.
  • The []PSNames checkbox (in Generate Font, Options dlg) did the opposite of what it was documented to do. Behavior has been changed to conform to the documentation (and expectations).
  • FontForge would crash when loading some sfd files from a few months ago.
• 26-Aug-2003
  • When I added metadata to svg fonts (20-Aug) I inadvertently put in two closing metadata tags. Which broke the font.
  • Fixed a bug conversion of context/chain coverage sub-table into a morx subtable.
  • Figured out in what cases I could generate a morx subtable from context/chain glyph or class format subtables.
  • Fixed problems reading required features from ttf files.
  • Updated French UI
• 25-Aug-2003
  • View->Show ATT didn't show contextual/chaining tables even if they would be put in the 'morx' table
  • View->Show ATT had problems showing anything in 'morx'
  • Make it possible to edit contextual/chaining lookups that are formatted by class
  • parsing/generating ttf files with class format context/chaining had problems
  • loading a class format context/chaining lookup caused memory problems when closing the file.
  • FontForge would sometimes lose track of script/languages of lookups when reading from a ttf file
  • Added a conversion routine from glyph to class format context/chain
  • Kanou points out that not all encodings available in Font Info->Encoding are present in the scripting interface, and fixes that.
• 22-Aug-2003
  • Fix simplify again
  • Patch from Andrey V. Panov to improve bluevalues
  • KDE refused to make the find problems explanation window visible, so now I shan't make it invisible in the first place.
  • Generate mac family crashed instead (introduced 2-July)
• 20-Aug-2003
  • Support for potrace
  • Fix some bugs with simplify more
  • Mac build was badly configured and linked with a shared library it should not have.
  • When compiled with -O2 fontforge had rounding errors which resulted in a bad bezier order3=>order2 conversion (PS->TTF)
  • SetPanose sometimes didn't
  • Patch from KANOU to set ulUnicodeRange for version 4
• 18-Aug-2003
  • FontForge did not interpret SVG relative paths properly, and did not generate correct SVG output because of this. FontForge was mistakenly assuming things should be interpreted as in the Type1 rrcurveto operator (where each point is relative to the previous point in the operator) rather than the rcurveto operator (where each point is relative to the initial point).
  • If FontForge opened a Bold font stored in OTF CFF format, then when it closed that font it would either crash, or screw up memory so that a crash would happen later.
• 17-Aug-2003
  • Support for contextual, chaining contextual and reverse chaining contextual sub-tables of GPOS/GSUB.
    • Will generate an Apple contextual glyph substitution table if:
      1. There is an apple feature which matches the otf tag
      2. The sub-table is in coverage format
      3. The sub-table contains either exactly one nested single glyph replacement substitution, or
         it contains exactly two single glyph replacements and one of them refers to the last glyph matched (and the other does not)
  • KANOU points out that on screens with high dpi FontForge would sometimes draw splines 2 pixels wide and sometimes 1 pixel wide (122dpi had this effect)
  • KANOU provided another patch for reading bitmaps one part covered a memory leak the other fixed a bug reading ttc files.
  • Added a LoadEncodingFile scripting command
  • New french ui.
  • FontForge was using the wrong definition of ascent & descent to set those fields in the 'hhea' table. This confused some mac programs (perhaps others).
  • Added a "BBox" item to CharInfo scripting.
  • Added a FontsInFile scripting command to return an array containing the fontnames of all fonts in a file. If an error occurs, a zero length array is returned.
  • Change so that when opening a file containing multiple fonts the font's filename contains the fontname (so it is possible to open other fonts from the same file).
  • Kanou provides a patch for compressing bitmaps with more than 8 bits of white space to their left.
  • Markus Schwarzenberg provides a patch to make fontforge recognize $cidfamilyname as it should have done.
  • Add the ability to generate fonts which contain both OpenType and Apple Advanced Typography tables (ie. a font that should work everywhere).
  • Added Utf8(int), Chr(int), Ord(int) to interconvert between strings and their encodings.
  • Add two problems to Find problems
    • Search for substitutions, contexts, kerning classes, etc. which refer to a name which does not match any glyph in the font
    • Search for contexts, chains, etc. which refer to a nested lookup which does not exist
  • Add a new command line argument "-last" which will load the last sfd file closed.

    ---

  • FontForge would crash when importing bitmap fonts
  • FontForge would crash when trying to do an Element->Default ATT->Suffix to Tag command
• 3-Aug-2003
  • I still wasn't getting r2l kerns correct
  • Add support for vertical kerning
    • In 'GPOS', 'kern' tables of Open/TrueType
    • In SVG fonts
    • menu commands
      • Remove VKern
      • Vertical Kern By Classes
      • VKern From HKern
    • scripting
      • RemoveVKern
      • SetVKern
      • VKernFromHKern
      • Add a "VKern" item to CharInfo
      • Type argument to SelectByATT has changed!!!
  • Add support for other Pairwise positioning features.
  • Add a $panose command to return an array of panose values (scripting)
  • Add a SetPanose command allowing scripts to change panose values
  • FontForge was not rotating .vert glyphs properly
• 31-Jul-2003
  • Installed KANOU's fix for a buffer overrun
  • Installed KANOU's fix for compacted cid-keyed fonts
  • Fixed another bug with compacted cid-keyed fonts
  • Mathias Wollin points out that my pfm files don't have the correct kerning table format
  • Handles the case where a font has both a 'glyf' and a 'CFF ' table by asking user which to use
  • Various improvements to the goto dialog from James Crippen.
  • If the ttf "Apple" box got checked then it remained checked in subsequent Generate Fonts even though the [Options] dialog did not show it checked.
  • FontForge would crash if given an Anchor class with no marks.
  • Hmm... I notice that volt only uses format 2 coverage tables. I wonder if windows fails to handle format 1 tables...
• 30-Jul-2003
  • Add support for svg fonts (and add dependency on libxml2 for reading them in)
  • Add support for importing/exporting svg files (again libxml2 needed for import)
  • Add support for required features.
  • In quadratic fonts the ellipse tool did not draw ellipses
  • Add support for rectangles/ellipses drawn center out rather than just within the bounding box
  • Preserve the rect/ellipse tool settings across invocations
  • Put a license entry into the help menu
• 20-Jul-2003
  • made the default linux browser be mozilla rather than netscape.
  • If the metrics view were displaying an anchored pair (mark-base), and a character were typed into the text field at top before the anchored pair, then the display got very strange.
  • FontForge got confused when merging a font which contained multiple encodings for the same glyph
  • FontForge got confused if there were an lparen in a directory name (thought this meant a font collection).
• 15-Jul-2003
  • We weren't parsing any type3 fonts properly any more
  • Extend our ps interpreter to understand some of METAPOST's idioms better.
  • I used to call all characters which weren't in the encoding ".notdef" this could cause problems if a user forgot to name them. I now give them unique names starting with "NameMe". I still leave characters <256 called ".notdef" if the encoding calls for it, this may be a mistake...
  • Fix on 14-July to simplify was incomplete.
  • More fiddling with autohint
  • Added support for the new X11 opentype bitmap format.
• 14-Jul-2003 (Bastille Day!)
  • There are still cases where I'm getting the truetype point count wrong. If a truetype file had a control point that lay under an end point, then I would treat that as a "no control point" mark, and ignore the cp. But if I do that blindly I mess up the point count, so during font generation I must be prepared to put the cp back.
  • Kanou provided a patch to fix a bug in the underlying graphics library.
  • Add a popup menu to Show Att to allow those data to be saved (as utf8 text)
  • Fixed (I hope) right to left kerning
  • Added 'ccmp' entries by default for Hebrew
  • Fixed a couple of bugs with the new simplify more options
  • fontforge didn't compile on solaris. It now does.
• 12-Jul-2003
  • Remove weirdness in scripting interface to TeX font parameters
  • Kanou points out that the code for reading format 4 embedded (ttf) bitmaps was ambiguous and would behave incorrectly sometimes. He provided a patch.
  • FontForge used not to create ghost hints for curved splines. It does now.
  • FontForge did not find some of the hints needed for n021004l.pfb
  • New French UI
  • FontForge would sometimes crash when autotracing characters
  • Don't snap to points when moving control points
  • Directory browsing didn't work if the local encoding wasn't latin1 and the directory name contain non-ascii characters.
  • Add an entry to find problems to look for 'GSUB' entries to non-existent characters.
  • Expanded the simplify more command
  • The GPOS output routines got very unhappy if a duplicate glyph happened to have GPOS data. Fixed on both ends I hope.
  • The pcf reader was unhappy if
    • the glyph bb width was a multiple of 8 (we read an extra byte per line which made the result appear as garbage)
    • several glyphs had the same name.
  • the FontInfo dlg failed to parse the merge_with field of Anchor Classes properly, meaning that each time [OK] was pressed in fontinfo each anchor class name got a number prepended to it.
  • Change the way the merge_with field of anchor classes is handled by the UI so that it makes more sense. We used to display a number, now display the name of another anchor class with which it can be merged.
  • Added an arabic text sample, courtesy of Thomas Milo.
  • Upgraded to Unicode 4.0 (Note: libuninameslist has also been upgraded)
  • Added a new preference item CharCenterHighest to assist in accent placement.

    ---

  • Oh. This isn't documented, but each langsys table can only point to one (1) feature with a given tag. So if I have multiple 'mark' lookups, I can't have multiple 'mark' features, I must have one 'mark' feature with pointers to many lookups.
  • FontForge could crash when generating a tfm file from a CID keyed font
  • Change GDEF output so that the class def table isn't generated when it isn't needed (we used to generate a 0 length one, which would confuse the freetype-1 opentype code, and perhaps others).
  • Fixed a bug in autohint caused by negative stem widths.
• 4-July-2003
  • Added a page to the font info dlg which allows users to set TeX font parameters
  • Added a mechanism for specifying TeX charlists and extension characters
  • Added a mechanism for specifying TeX Italic Correction
  • Revamped support for reading tfm files so that it sets all these new things correctly.
• 2-July-2003
  • Add an Options dialog to the Generate Font window, and remove the check boxes from it.
    • Add a UI for the 'PfEd' table
    • Add a UI to turn off PostScript hinting
  • FontForge can now generate a tfm and enc file (for TeX)
• 30-June-2003
  • In the table of contents at the start of a true/open type file the GDEF table was listed as having the wrong size. This caused many applications to ignore it, which in turn meant that Mark to Base GPOS tables were ignored (at least I hope that's what was going on).
  • On the mac local help files weren't working properly.
  • Combining unicode characters (mark glyphs) should have an advance width of 0, and now they default to that.
  • Several people at EuroTeX wanted more pen shapes in the expand stroke dlg. Ellipses are easy, so we add them.
  • One guy at EuroTeX wanted to be able to store arbitrary per-glyph data in the font, so add a new (private) true/open type table that can contain the glyph comments. This table has tag 'PfEd' and can contain sub-tables containing font comment, glyph comments, glyph colors. Perhaps other data as need arises.
  • Kanou has added a new scripting command, SizeOf, to return the number of elements in an array.
• 19-June-2003
  • Added a histogram dialog to help figure values for StemSnap[H/V]
• 17-June-2003
  • Scripting interface to SelectByATT
  • Crash bug caused by incorrect CID fix on 15-June
  • (the web interface to the cvs tree seems broken. It doesn't show rmligamarks). Here's a copy of rmligamarks.c.
• 15-June-2003
  • Cleaned up the spelling of some scripts.
  • Deactive the [Edit] button of the Script/Lang list when appropriate.
  • FontForge wasn't removing the quotes around the copyright notice when read from a bdf file. After several rounds of read/save there were lots of quotes.
  • Added a wireframe effect.
  • Added a find problems entry which will check for control points which are very close to their main point (so close that it is unlikely that they will effect the curve much).
  • The Undo command from the font view did not reset the character changed bit.
  • GDEF table output had problems with cid-keyed fonts.
  • So did 'prop' table output.
  • Expanded View->Show ATT to show 'GDEF', 'lcar', 'prop', 'opbd' tables.
  • Added a Ligatures entry for View->Combinations in the outline character view
  • Added "Edit->Select->Select by ATT", to allow user to select characters depending on advanced typographic features.
  • FontForge always used to set the "ignore-combining-marks" flag bit for all ligatures. But yudit (and probably gedit) will not process the substitution if that bit is set. On the other hand I believe some arabic ligatures may depend on it (at least arabtype-volt.ttf uses it). So... now we only set this bit for arabic ligatures which do not contain combining marks. Dunno if that is right either.
    • I have created a little tool, rmligamarks, which will take an sfd file with the "ignore combining marks" bit in ligatures and turn it off. It does not try to guess what should be done, it ALWAYS turns the bit off.
• 5-June-2003
  • FontForge was generating Apple 'morx' tables with the right to left bit reversed (so it thought latin text was right to left and arabic was left to right). This meant that ligatures happened backwards (the "ffi" ligature would be invoked if you typed in "iff")
  • FontForge was generating a format=1 glyph class definition sub-table of GDEF. ttx didn't like this, I'm not sure why. Problem goes away if I generate a format=2 sub-table.
• 4-Jun-2003
  • Added a Shadow effect
  • More fixes for remove overlap
  • fixed a bug with timers
  • Preserve the Fit To Em mode of a font.
  • Added a bunch of scripts (as in writing systems, not executable files) from the proposed ISO 15924 standard.
  • Added support for underline position to bdf files.
  • The 18-May build was bad and failed to read postscript fonts properly on many platforms (the sources were ok, but the builds were bad)
  • Display the version number in the about box, and attach it to the directory in the source .tgz file.
• 18-May-2003
  • FontForge would crash when asked to produce a ttf font where there was an unused anchor class.
  • The original encoding is now preserved when a font is generated.
  • The goto dialog (and scripting) now recognize "glyph<nnn>" as refering to glyph <nnn> in the original encoding.
  • Kanou provided two scripting commands, ConvertToCID and ConvertByCMap.
  • Added a Remove External flag to Expand Stroke to complement Kanou's Remove Internal
  • Expanded on Remove Overlap by adding
    • Intersection -- which removes everything except the overlap
    • Exclude -- which removes areas where any selected contours cross non-selected contours (and then removes the selected contours)
  • Added an Effects sub-menu of Elements, including (so far)
    • Outline -- turns the selected characters into outline versions of themselves
    • Inline -- turns the selected characters into inline versions of themselves
  • FontForge would crash when reading in an sfd file containing a kerning class
  • gimp objected to some of my ttf fonts. I think because it is inappropriately clipping to the ascent/descent values in the hhea rather than using the font's bounding box. It's easy enough to preserve the 'hhea' ascent/descent values and perhaps that will be enough to make gimp happy.
  • Kanou points out that rasterization has problems with the new TOfNextMajor, so use the new code for remove overlap and friends and the old code for rasterization
• 12-May-2003
  • Added a command to do non-linear transformations (but made it require user configuration to have it compiled)
  • Added a UI for Kanou's Remove Internal flag to Expand stroke
• 9-May-2003
  • Apple's 'lcar' (ligature caret) table was being generated incorrectly
  • Another bug in point numbering
  • Make "Slanted" a synonym for Oblique.
• 8-May-2003
  • FontForge did not recognize "filename(fontname)" notation for loading a specific font from a resource fork of a mac file.
  • FontForge crashed on the mac when generating a GPOS table. There was an uninitialized variable there, left behind be recent changes. Removed.
• 7-May-2003
  • Oops. It should be legal for a 'mkmk' anchor class to have two points in a glyph, one the normal mark point, the other the base mark point. (that is it should be possible for a glyph to attach to itself).
  • Kanou supplied a patch to extend expand stroke.
  • Changed behavior of AnchorClasses so that two classes could be associated to form a single feature (ie. one class of "Above" marks, and one class of "Below" marks)
  • Show ATT now displays something meaningful for marks
  • Some general Anchor Class clean up.
  • Was not setting bits in the CodePage[1] field of the OS/2 table for all the weird MS specific encodings. It seems I should do that.
• 6-May-2003
  • Although undocumented, digits behave just like lower case letters in macintosh filenames for postscript resource fonts.

    ---

  • Looking at the font info->ttf values dialog would cause fontforge to crash after [Ok] was pressed.
  • Old sfd files with anchor classes but not script lang index would crash fontforge when being read in.
• 5-May-2003
  • Made it possible to select kerning class 0 (ie. everything that isn't specified)
  • Put in a check to make sure the user didn't add a glyph to two different kerning classes
  • FontForge was dumping out a full GSUB sub-table for each character that sub-table applied to (rather than one sub-table for all characters). So there were tons of duplicates.
  • FontForge was producing extension sub-tables that held an incorrect offset.
  • Added some code to protect FontForge when it read incorrect ttf fonts like those generated above.
  • Drag and Drop to the Char Info dlg was broken.
  • If a font had a non-notdef character at encoding 0 and we used freetype to generate bitmaps for it, then those bitmaps would be assigned to the wrong glyphs (ie. the bitmap for "A" would go to "B", etc.)
  • I used "Mac Roman" in several places where I should have said "Macintosh Latin"
• 4-May-2003
  • Added code to determine better defaults for the script lang index
  • The ligature state table in Apple Chancery has so many possible paths through it that FontForge bogs down when evaluating it. I presume it would finish in finite time, but it takes unacceptably long. So I've put in an abort.
  • Add support for KernClasses (ie. they are no longer just converted into a large number of kernpairs, but will be stored internally as a set class and will be output as such). And they have their own dialogs to manage them.
  • Add support for Apple's new version of the 'kern' table.
  • Add support for the format 2 sub table of Apple's 'kern' table (which is different from OpenType's format 2 sub-table of the kern table), and is Apple's kerning class sub-table (note: Apple's docs are WRONG. they claim class data are stored in an array of uint8, but they really are in an array of uint16).
• 29-Apr-2003
  • Oops, one more fix to remove overlap
  • Kanou points out that copy hasn't been working (for some time now) when a contour has be split into bits, only one of the bits is copied.
  • Kanou points out that Paste reorders open contours
  • Further work on Show ATT dlg
  • Further work on multiple features.
• 28-Apr-2003
  • Yet more fixes for Remove overlap
  • Added a mode to the ruler tool where it shows the exact position of the mouse cursor
  • The Goto command had a strange bug: If a font had a glyph with a known name, but that glyph did not reside at the unicode slot assigned to that name, then goingto that name would go to the (empty) unicode slot, rather than to the actually character.
  • Werner wants a dlg which shows collects all the GPOS/GSUB (morx/kern) information in one place rather than scattering it around hundreds of dlgs.
  • FontForge now stores the script/langs for kern pairs
  • FontForge failed to read script/langs for GPOS properly.
  • FontForge had problems generating an AFM file from a CID font with ligatures
  • FontForge had problems getting script lang lists right when multiple features refered to the same lookup.
• 20-Apr-2003
  • One of FontForge's deficiencies is that it handles multiple encodings for the same glyph very badly (it does not have the multiple layers of indirection that fonts do). Rather than solve this properly I have added a new command that allows for this case. SameGlyphAs.
  • Add some more synonyms for greek glyph names
  • Failed to realize that U+1FED was an accent
  • No longer tries to build greek accents out of latin ones when Build Accented is invoked. (but still will for Build Composite).
  • When swapping character names, check for names like "uniXXXX" and instead of swapping it, rebuild it as appropriate for the other character.
  • Font Info->Encoding failed to change the encoding properly when switching from "Compacted" to anything other than the original encoding.
  • A couple of changes to the italian translation
  • Added an "Original" encoding which gives you back the ordering of the glyphs in the ttf font. (this may be changed when you regenerate a ttf font)
  • FontForge failed to read ttf instructions for composite glyphs.
  • FontForge got confused if a ttf file had multiple glyphs named ".notdef" (which to me looks like a bug in the ttf font, but Times New Roman does this so I have to deal with it).
  • Added a command to the popup menu in the charview to bring up a window looking at a reference.
  • Add a check for when sfd files contain bad kerning info.
• 14-Apr-2003
  • Change a tab in Font Info from "PS General" to "General"
  • When searching for encoding files, look for the extension ".enc" used by TeX for PostScript encoding files
  • Fixed a crash when setting a character's name to a name already given to another character
  • Added many new X resources for controlling colors in the outline character view.
• 7-Apr-2003
  • Program generated divide by zero errors when using Meta-Shift to move a control point which didn't exist.
  • (cleaned up some syntax issues in non-English nomen files)
  • Added new scripting commands ClearPrivateEntry, ChangePrivateEntry, GetPrivateEntry
  • Added new scripting command: UnicodeFromName
  • Added new script command: RemoveATT
  • Vadim pointed out that loading "pala.ttf" and then generating it again (as ttf) produced bad results. FontForge was getting confused by the points which were used to control Anchors, getting its points list out of order and generally screwing up.
• 04-Apr-2003
  • autosave sometimes lost track of which font it should be applying changes to.t
  • Add Extrema (or just adding a point in the middle of a spline) would generate incorrect splines if the spline were linear in one but not both of the two dimensions
  • One more change to Merge/Simplify. Deal with the case of a curve point with no control point (use the control point in the opposite direction to get the slope).
  • Typo in 2-Apr script/lang fix caused crashes.
  • TeXBaseEncoding (T1) changed in Oct-2002. Updated my version (also updated the other TeX encodings I've got and included the cyrillic ones into Encodings.ps.gz)
  • Sometimes we failed to remove the dot when creating a dotless i/j character.
  • Merge/Simplify had conniptions when given a zero length spline.
  • Added "DefaultATT" and "AddATT" scripting command to allow user to add Advanced Typography features to characters.
• 2-Apr-2003
  • Problems with script/lang when: Pasting or Merging ligatures and substitutions from one font to another
  • FontForge could generate points (within a ttf simple glyph) in a different than the original glyph, meaning that the instructions would move the wrong point resulting in blots. I have fixed this (or at least the cases I found).
  • FontForge would always give ASCII characters their standard names in a ttf post table. This was wrong if the character did not have its standard name.
  • FontForge could go into an infinite loop on some simplifies.
• 1-Apr-2003 (April Fish!)
  • Fixed a crash in the Join command
  • Fixed a crash when loading some truetype fonts (never bothered to figure which ones exactly)
  • Fixed a similar crash when loading old sfd files (note some old fonts won't guess the script for ligatures, etc. properly)
  • Added a new preference item: Stop At Join
  • Changed the way control points are set when changing point type.
  • The fill of the outline view sometimes lagged the change.
  • Added another tag to CharInfo scripting command: GlyphIndex. It returns the glyph number in the ttf font.
  • Werner suggests that instead of naming things "nounicode-%x" where %x is the encoding, use "nounicode-%d-%d-%x" where the two new arguments are the platform/specific pair used in the cmap we picked.
• 31-Mar-2003
  • FontForge was not reading pfb fonts properly if they had more than one binary section header.
  • When FontForge removed a character it did not remove all the kernpairs that refered to that character.
  • Fixed more leaks with valgrind.
  • Added more synonyms for glyph names from Adobe Glyph List (I still rely on Glyph names for New Fonts as my basic set of names)
  • FontForge had problems if the use selected 0 levels of undoes when the UI was active.
  • Selecting the freehand tool in a quadratic font caused a crash. (The freehand tool doesn't work in quadratic fonts, but it shouldn't crash)
  • Changed the way scripts are specified. I used to assign a script to every character. I realize that doesn't work. The digits are used in many scripts and the fraction ligatures based on them need to be seen in all those scripts, not just the one that the digit was assigned to. So...
    • Glyphs no longer have a script assigned to them
    • Substitutions, Ligatures, Anchor Classes and other Advanced Typographic features now have a list of scripts (and languages) for which they are active.
    • There is now an absurdly complex stream of dialogs which may be used to set this list.
  • Added a new preference item to allow the user to specify a local help directory
  • Added a new scripting command SelectIf which allows you to determine whether a character is in a font (Select signals an error if the character isn't there, SelectIf returns failure).
  • Added a new scripting command to change a font's order (from quadratic to cubic)
  • ScaleToEm scripting command will now accept a single argument, the em-size.
  • Added a new scripting command to allow you to apply an otf substitution to the font
  • Added a new scripting command to give control over which ligatures are output in an afm file
    • (it used to be that all ligatures were output in the afm file, no only 'liga' and 'rlig' that appear in a default language are output)
  • AFM files only support 2 character ligatures. So, when possible, generate a sort of closure on the ligatures. So if we are given the two ligs "ffi->f f i" and "ff->f f" then add the ligature "ffi->ff i". This means that the sequence f+f->ff+i->ffi will eventually reach the ligature we want.
  • Changed so that an untitled anchor class is now called "Anchor-*" rather than "Untitled-*".
  • Rewrote Merge & Simplify commands. In my tests they now produce better results.
  • Type2 font generation did not make use of hv/vh curveto commands in some cases
  • When figuring out whether to synchronize the width/lbearing of a base character with the composites containing it, I only checked if the base character was alphabetic. But that doesn't work for hebrew where unicode says the marks are alphabetic. So add an additional check that it can't be a combining character.
  • When pasting a composite character from one font to another, the saved width is not useful. Added a check that finds the correct width of the appropriate base character.
  • When pasting a composite character from a font with one em-size to another with a different em-size try to scale the amount references are translated by an appropriate amount.
  • CharInfo scripting command has been extended to allow users to test the presence of substitutions
  • Some fonts (bsmi00lp.ttf for one) have bad cmap tables which map several glyphs to the same unicode code point. Put in a check for this and only map one glyph to that code point now.
• 13-Mar-2003
  • Add a preference item to control the number of undoes saved per character
  • Turn off saving undoes when in a script started from the command line
  • Added a preference item to control the undo depth from the UI.
  • Try to do better about scaling my chunk allocator to machines with 8 byte pointers
  • Was doing 32 bit arithmetic when reading GSUB simple substitution lookups when I should have done 16 bit.
  • Tracked down a lot of memory leaks with valgrind (and some uninitialized variables too)
  • Added access to a font's version from scripting
  • Added PostNotice scripting command
  • When reading a ttf/otf font with a GSUB table fontforge will try to guess reasonable names for unnamed glyphs by looking them up in the GSUB table. Thus an unnamed vertical variant for "A" might be named "A.vrt2".
  • FontForge would crash when outputting ttf files with no 'cvt' table (bug introduced around 27-Feb)
  • the configure script did not respect the argument --with-freetype-src as it should have.
• 10-Mar-2003
  • Added a TrueType instruction debugger (actually an interface to FreeType's truetype bytecode interpreter)

    This command depend on your compiling freetype with the byte code interpreter enabled. This is disabled by default because it infringes on certain patents granted to Apple. If you have a license from Apple then you may enable the interpreter by setting the appropriate macro in .../include/freetype/config/ftoption.h before you build the library (see the README.UNX file on the top level of the freetype distribution). FontForge needs to have the freetype source directories available when it is built (there are some include files there which it depends on)

  • Added an Edit->Copy Grid Fit command (to put the grid fit splines into the clipboard)
  • Added a warning if the freetype library doesn't have a byte code interpreter so that the user knows his/her splines have been autohinted (rather than had the bytecode instructions run).
  • Improved Edit->Merge in some cases that don't fit its constraints
  • KANOU found some more underscores in character names
  • And provided a new japanese ui
  • Round to Int had problems on quadratic splines
  • Added support for runtime loading of dynamic libraries on the mac.
• 03-Mar-2003
  • Memory got screwed up when converting TTF splines to PS ones
  • KANOU suggested some changes to the quadratic spline UI
  • I had the wrong makefile at top level so builds didn't
  • syntax error on systems without freetype
• 02-Mar-2003
  • Changed the Show Instructions command to be an Edit Instructions
  • Bug saving 'prep' table
  • Bug changing 'cvt ' table
  • Added ability to look at gridfit splines in charview.
• 28-Feb-2003
  • New dialog in outline view to examine ttf instructions
  • If an sfd file was created from a script then when it was next opened from the UI it would appear in a tiny window.
  • The preference item refering to the []Hint check box in the generate dlg has been renamed.
  • A checksum-like thing has been added to ui files so that in the future the application will refuse to use .ui files which do not match its expectations.
• 27-Feb-2003
  • Didn't get the SetLBearing change right yesterday.
  • Changing to quadratic splines with Font Info could generate an invalid spline.
  • We now preserve truetype instructions when reading in a truetype file (the fpgm and prep tables, the cvt table and any per character instructions)
  • These instructions will be dumped out when generating truetype fonts
  • Added a menu command AutoInstr to generate our own (bad) instructions for a character.
  • Any significant change (ie. anything that reorders the points, or adds or removes points) will clear the instructions for that character.
  • Changing to cubic splines (fontwide) will clear the instructions (and the fpgm and prep tables).
• 26-Feb-2003
  • configure scripts got confused in the freetype libraries were findable but the include files were not. Changed the scripts so that freetype is marked not present if the include files are missing.
  • Table ordering was not read in properly from an sfd file
  • Typos in SetCharName & SetUnicodeValue prevented second argument from working
  • Added a second argument to SetLBearing and friends to allow relative bearing settings (ie. to increment the bearing by some amount)
• 25-Feb-2003
  • Bad bug in GSUB generation introduced yesterday
  • Font level round 2 int command now rounds hints too.
  • Adobe lies when they say that a hint can't stretch from one bluezone to another, so be a bit more cautious in removing this case.
  • Be a little better about getting the script from the GSUB/GPOS table.

    ---

  • FontForge had problems generating mac cursive connection setting.
  • FontForge had problems defaulting ligatures in cid keyed fonts (even if there were no ligature characters to be defaulted, the attempt would fail)
  • I introduced memory problems yesterday. Fixed.
  • Font level round 2 int will take hints into account when positioning points in order to maintain stem widths.

    ---

  • When running a script from the command line FontForge failed to initialize the adobe standard encoding.
• 24-Feb-2003
  • Claudio Beccari has provided an Italian UI.
  • (added Italian style translations to fontinfo)
  • User now has the ability to control the conversion from mac feature/settings to otf features.
  • FontForge now remembers the order features are to be applied in GSUB/GPOS/morx tables (and dumps features out in that order when it next saves them).
  • FontForge allows you to reorder the GSUB/morx tables should you wish to do so. I am sufficiently ignorant that I don't believe it is necessary to reorder the GPOS table, so there is currently no UI for this. If I am wrong let me know and I will add that too.
  • The 'kern' table was not being generated when it should have been. This error was introduced sometime in the last month due to the AAT changes.
  • Scripts generating otf fonts would crash if the font had non-integral coordinates because of a bug in my PostNotice routine when running without X.
• 18-Feb-2003
  • Hmm. It looks as though apple uses the "ascending glyph" flag on the features of the 'mort'/'morx' tables to mean left to right text
  • Fixed a crash on reading a GSUB table with a glyph out of range.
  • Added support for a cursive connection (contextual glyph substitution subtable of) morx.
  • Added an index to this manual and put a link to it in FontForge's Help menu. Index is by no means complete yet.
• 12-Feb-2003
  • Support direct editing of quadratic (truetype) splines
• 11-Feb-2003
  • Change made on 8 Feb 2003 broke the code that read type1 flex hints
  • Hide files beginning with "." (except for "..") in file chooser dlgs.
  • Add protections in the afm file output to handle very long copyright notices.
• 08-Feb-2003
  • Support for Apple's 'opbd' (Optical Bounds) table
  • Support for reading some features in non-contextual glyph and ligature sub-tables of the 'mort' or 'morx' tables.
  • Support for writing those features in non-contextual glyph and ligature sub-tables of 'morx'.
  • added char info to the fontview popup.
  • the scripting command Generate (and GenerateFamily) would generate truetype .dfont files when asked to generate otf dfont files.
  • Add ClearHints scripting command
  • Add access to preference items from scripting (via: $<prefname>, GetPref("<prefname>"), SetPref("<prefname>",<value>)
• 01-Feb-2003
  • Add some support for Apple Advanced Typography tables ('prop', 'lcar')
  • Popup menus in textfields failed to refresh the text field
  • KANOU provides a patch for sfd files with long Alternate Subs
  • FontView didn't deselect properly after doing a drag and drop.
  • Should now deal with the case where (under cygwin) the user sets BROWSER to 'netscape' with no path spec.
  • Ligature carets introduced a crash bug.
• 30-Jan-2003
  • Should figure out the default windows browser under cygwin now.
  • Should handle the case where a user sets the BROWSER variable to something with a DOS pathspec
  • Apply KANOU's patch to turn off more autohinting.
• 29-Jan-2003
  • Fix a crash introduced yesterday.
  • scripting didn't work for $familyname fix from Alex Raftis
• 28-Jan-2003
  • The title of the Anchor Pair dialog was wrong (said Kern Pairs)
  • Improved error messages for when a font file did not exist or was unreadable.
  • Failed to generate a GSUB table when the ligatures came from a r2l ttf file.
  • I was generating incorrect data for r2l GPOS kerning sub-tables.
  • Failed to display r2l anchored attachments properly in the metricsview, anchor pair dlg, or when printing.
  • Typo when loading libuninameslist dynamic library, introduced on the 24th.
  • Now we generate a GDEF table too, containing (possibly) a glyph class def and (possibly) a ligature caret list
  • Added code for editing horizontal ligature caret locations
• 25-Jan-2003
  • More cygwin fixes for invoking browsers.
• 24-Jan-2003
  • Added a preference item to control whether autohinting happens before a font is generated.
  • When calling Internet Explorer or Netscape on cygwin we needed double backslashes
  • moved hidden menu items (those invoked by Shift) into submenus
• 20-Jan-2003
  • Put in a number of errors and warnings related to a glyph not having a script.
  • Fixed a crash bug when generating a popup for a glyph with no unicode encoding
  • Support for scroll wheels.
  • I hope I've now fixed the problems related to the GSUB table and ttc files. Now does a pass through the GSUB table to see what glyphs can be introduced by it when it figures out what glyphs to read from the glyph table (ttc files have more than one font sharing the table, don't want to read all glyphs).
  • Added support for extension subtables of GPOS/GSUB.
  • Making a selection in the textfield of the Display dlg was broken when the selection was scrolled.
  • Valek wants a Paste Into command.
  • Uninitialized variable in GSUB output caused erratic ligature behavior
  • I'm told that even though it makes logical sense for hebrew to have a 'fina' tag, it is not customary and I have been asked to remove it from Default ATT. Which I have done.

    ---

  • FontForge now protects itself better against erroneous GSUB/GPOS tables.
  • Valek points out there were display problems with the File->Display (and File->Print) menu options in Russian.
• 16-Jan-2003
  • Recent char info changes broke copy & paste within text fields of the char info dlg.
  • Problems reading GSUB table of a ttc files (problems remain, but we should no longer crash).
  • Reorganized libraries. Now FontForge has an additional dependency:
    libuninameslist. Please install this.
  • With libuninameslist FontForge can display the Unicode annotation information.
• 14-Jan-2003
  • Fixed a couple of bugs loading pcf files with complex encodings
  • Bug in pasting metadata when there was no glyph comment.
  • Made SetCharColor scripting command set the color on any selected glyphs
  • Cleaned up selection behavior in fontview. Drag & drop made it difficult
  • Oops. Was scaling round rect radius incorrectly in xfig conversion
  • Oh yes, hebrew has final letters too. And greek. Add them in to the default for 'fina'.
  • Fixed many problems with mac font generation
  • Open VMS update from Jauk Jansen
• 12-Jan-2003
  • Support for xkb extension
  • New Russian UI from Valek
  • New Japanese UI from KANOU

    ---

  • Saving a compacted font did not preserve the old encoding
  • Added a preference item to control Edit->Join
  • made Edit->Copy From->Copy Name (now Metadata) be a preference item too.
• 9-Jan-2003
  • Support for round rects in xfig files
  • Added the glyph name to the title of the char info dlg
  • Char info failed to deal with all the flags properly
  • Fixed crash when creating GPOS sub-table 1 (simple positioning)
  • GPOS/GSUB output now handles case where a feature has several different sets of flags.
• 8-Jan-2003
  • Update copyright notices
  • Printing now handles anchored marks. (Display does not, just as display doesn't handle kerning-- because X doesn't).
  • Support for cursive positioning.
  • I've registered with Adobe, and they have assigned FontForge an XUID base number of 1021. Now when new users start fontforge for the first time they will be assigned an XUID base sequence of 1021 xxx nnn where xxx and nnn are two random numbers generated by fontforge. New fonts will get a fourth number for their full XUID.
  • Kanou points out that glyph names should not contain underscores so redo the .cidmap files to change to dashes. Also change "japan1_%d" to "cid-%d" to make life easier for me.
  • KANOU points out there were places where fontforge did not check psunicodenames_cnt. Once outside of BMP this causes a crash.
  • New Japanese UI from KANOU.
  • Support for many more GPOS/GSUB sub-tables.
  • Keep track of the order in which characters are selected in the fontview and recreate that order when user asks to create a metricsview based on the fontview's selection
  • Make it possible to drag&drop characters from fontview into metricsview (again keeping track of selection order)
  • Make it possible to drag&drop characters from fontview into charview to create references
  • Make it possible to drag&drop characters from fontview into charinfo to create ligatures, etc.
  • Command to set the default values for sub GSUB/GPOS features that are easy for FontForge to figure out.
  • Ability to zoom metrics view.
  • Metrics view displays anchor points and does the needed work to position the characters. It is also possible to move anchor points around in the metrics view.
  • Metrics view now has a pop-up menu that allows you to see alternate metrics
  • Metrics view now has a submenu (off of View) which allows you to change the current glyph to one of its alternates.
  • Fixed uninitialized variable what would sometimes cause a crash when storing bitmaps in a TTF/OTF file.
  • Fixed crash bug in 29-Dec when generating a truetype font without a GPOS table.
  • Many changes to charinfo dlg. Made non-modal. Can Copy/Paste Position/Substitution data.
  • Do a minimal amount of work toward supporting the alpha channel of a png file.
• 29-Dec-2002
  • FontForge now supports anchor points for mark to base (ligature and mark) positioning.
  • FontForge was mistakenly treating NumLock on PC keyboards as if it were the command key on Mac keyboards.
  • The font info dlg now remembers what tab was last pressed.
  • The char info dlg has been restructured. It includes a script field, and associates a tag with the ligature (if any).
  • I have decided to generate a 'GPOS' table rather than a 'kern' table in truetype output as well as in opentype output (unless applemode is set).
  • I have changed the default names of items in adobe's private use area to match their new conventions. ie. names used to look like "Asmall" and now look like "a.sc".
  • I have rewritten the way GPOS and GSUB tables are output. I hope this will not have much effect on users, but it does mean that a few more sub-tables are now supported.
• 23-Dec-2002
  • Various improvements to cubic->quadratic (truetype) conversion.
  • When switching from displaying a 12,8,6,4,3,2,1pt bitmap to a 24pt anti-aliased version, fontforge would fail to update the display.
  • The bitmap dlg would come up with the labels for X showing when it should have shown those for Win/Mac.
  • Inspired by KANOU's mentioning the mona font, I've made it possible to generate an ms bitmap only ttf font. This is done by creating dummy glyph and loca tables (full of space glyphs) and putting a bitmap scaling table which causes the most likely pixelsizes to be scaled from the strikes available.

    You can generate this with: [No Outline Font] [In TTF]

  • Drat. Fix of 24-Nov was wrong. I still wasn't putting the UniqueID into the Private dict.
  • I've been generating bad macbinary files since I started
    • I had not noticed that the resource and data fork must be padded to a multiple of 128 bytes.
      • This also meant that on rare occasions I would read them incorrectly too.
    • I was generating the crc incorrectly.
  • Be a bit more vigorous about displaying filter text in file choosers.
  • Default buttons didn't respond to the keypad enter key.
  • Pressing OK in the character info dlg would often destroy the font's encoding event when it should not have.
  • Was generating bad kerning tables for mac FONDs.
  • Jim Killock reminds me that there was no way for fontforge to read kerning data from a mac FOND. File->Merge Kern Data will now accept a mac resource file.
• 15-Dec-2002
  • Add support for Big5HKSCS
  • Redo CJK encodings
  • Update cidmap files.
• 14-Dec-2002
  • Add support for reading windows .FON and .FNT files.
  • Add support for writing windows .FNT files. Sadly .FNT files aren't particularly useful. They need to be converted to .FON files with the resource editor somehow.
  • Bug when reading postscript files that included complicated combinations of translate scale and concat commands.
  • FontForge was a bit too cavalier about guessing how big serifs were. Caused problems when autokerning a fraktur font. (FontForge thought it had huge serifs when it had none).
  • Direction (Clockwise, counterclockwise, correct) menu items are now greyed out if the selected path is open.
  • New OpenVms patches from Jacob Jansen.
  • Update unicode database to 3.2
• 13-Dec-2002
  • FontForge will now open a ttf/otf font that contains only bitmap strikes
  • FontForge will now read NFNT resources (old mac bitmaps)
  • FontForge will import NFNT resources
  • Fixed a number of problems with loading and saving bdf files
    • Did not understand that the "ENCODING" keyword could take two values if the first were -1
    • Added unencoded glyphs at the wrong encoding (I know that sounds like an oxymoron. I assign them an encoding, and I did it badly).
    • If a font name were not one of the X Logical Font Names then fontforge might fail to give the font a familyname or a fontname. This caused more problems when generating a new bdf file.
• 11-Dec-2002
  • Revert Glyph is now undoable
  • Fixed a crash when pressing return in the font display of a bitmap font.
  • Display the name of references near the top of the reference in the outline char view.
  • Took the width display off the info line and did it as hints are displayed.
  • Menus now respond to Home and End key events.
  • I have changed the calling conventions for the ExpandStroke scripting command to reflect the height ratio value that is now needed.
  • Holding down the Alt (Meta) key when using the freehand tool will now close the path.
  • The Generate Fonts dlg will now display an appropriate bitmap fontname if "No Outline Font" is selected.
  • I think I've got the wacom stuff working now. FreeHand tool now takes options.
• 08-Dec-2002
  • Oops. Can't use Alt-Shift for menus (it's treated as a mnemonic) use Alt-Ctl-Shift instead.
  • Reorganized the prefs dlg.
  • Made FontInfo non-modal.
  • Added a default icon for all fontforge's windows & dlgs.
• 06-Dec-2002
  • Fixed a problem where FontForge still picked bad fonts sometimes.
  • Typo in format spec for character windows
  • At Werner's suggestion:
    • The Transform dlg now preserves hints in the undo
    • Added an indication of which layer was active
    • Added an indication on the title bar that a character was changed
    • Added a Revert Glyph command
    • The location of the advance width pops up on the info line when the cursor moves over the advance width
    • changed the Set Width command to get the right default value.
    • There were some layout problems in the char info dlg.
  • Redid the configuration for mac os/x. I've just upgraded to the new developers' tools and that fixed a lot of problems.
• 05-Dec-2002
  • Fixed a crash in menus, if the cursor was moved from a submenu (a cascaded menu) to the menubar without passing through the main menu then a crash would happen.
  • Clicking in a menu above the first line will dismiss the menu.
  • Fixed a crash when generating postscript fonts.
  • Oops. Bug in work of 2-Dec. The character named "u" was treated as uXXXX and given unicode U+0000 rather than U+0075.
  • It is now possible to switch between editing layers in the outline character view by using the mnemonics displayed in the layers palette (even if the layers palette is hidden).

    Or by using the popup menu invoked by the third mouse button.

  • Rewrote Expand Stroke so that the calligraphic pen is no longer treated as a line (it is now treated as a rectangle).
  • Changed behavior of Next/Prev char so that they do not wrap in the outline character view. Also added the encoding to the title.
  • In the outline view made a separate menu entry for Char Info and removed it from "Get Info". Made Ctl-Shft-I invoke Char Info in all views (Ctl-I will also invoke it everywhere but the outline view).
  • Undo will now attempt to reset the modified bit on a character. It will not attempt to reset the modified bit on the font as a whole (there are too many things that can change without setting undo info).
• 2-Dec-2002
  • If the fontview were only one line high FontForge would not scroll to the selected character properly.
  • Adobe has released a new version of their glyph naming conventions. In particular they provide names for code points outside of the bmp. Danilo Sagan points out that they also allow for alternate glyph names. (specified in the FDK).
  • Remove the PFAEDIT_PI1 environment variable.
  • Cut and Paste bug fixed when clearing metadata.
• 25-Nov-2002
  • Harald Harders provides a patch to allow SetItalicAngle to deal with real numbers.
  • Vadim Belman provides a patch to keep Element->MetaFont from crashing when applied to spaces.
  • Bringing up the preference dlg would segfault if there were arguments specified for autotrace.
  • Put in a hack so that the environment variable PFAEDIT_PI1 controls whether U+03D6 is named "pi1" (as Unicode suggests it ought to be) or "omega1" (as Adobe has said it should be).
  • Follow riggle's suggestion and use the shell command "open" (on a Mac) to bring up a browser window.
• 24-Nov-2002
  • Harald Harders points out that UniqueID needs to be in both the font and the Private dictionaries in a PostScript font.
  • And provides access to UniqueID from the scripting language.
  • More fixes for TeX sfd files.
• 20-Nov-2002
  • Changed the Simplify More command to bring up a dlg giving the user control over what errors Simplify More is allowed to induce.
  • Bug in import image with non-latin1 fontnames
  • New Russian ui translation
  • Merge fonts didn't work well with CID keyed fonts
  • Fixed yet another problem with notdef generation.
  • Fixed yet crash in sfddiff.
  • Using Font Info to convert to SJIS didn't work.
  • Merge fonts had problems when both fonts were CID keyed.
  • Generate a unicode encoding subtable even when generating a CJK encoding subtable.
  • Added a preference item to allow user to specify whether s/he would rather read in the unicode or the cjk encoding subtable.
• 13-Nov-2002
  • Two new tools in the outline character view.
    • A hand tool -- for scrolling around
    • A freehand tool -- for having fontforge try to fit splines to a hand drawn curve
  • MergeFonts was broken from scripts.
  • New Russian UI from Valek.
  • Another remove overlap fix from KANOU.
  • Added code to invoke Internet Explorer (or an Mac App) on the Mac (so pressing F1 should actually load some help on the mac now).
  • Fixed random crash in font generation dialog
  • PS->TTF conversion tried to force splines in cases where it could not.
• 05-Nov-2002
  • Improved behavior of Remove Overlap. Fix inspired by Kanou.
  • Added a "Find Intersections" command, a variant of remove overlap that stops after it finds the intersections. I think this half of the job has no major bugs.
  • If a single point is in the clipboard, then FontForge will also export the clipboard as STRING, giving the coordinates of the point. This is to make it easier to refer to the point.
• 04-Nov-2002
  • KDE has the lovely habit of losing some resize events when in continuous resize mode. So I've tweaked the fontview resize code so that it is less dependent on individual events.
  • Added a mode where the user can choose between whether the font shown in the fontview should be scaled so that the em fits in the space provided (this is the old behavior) or so that the bounding box does.
  • Rewrote the knife tool so that it cuts at the end rather than as it goes along. I think this is easier to use.
  • Bug in the generation of format=2 truetype encoding tables (used for CJK fonts).
• 02-Nov-2002
  • Png reading failed on greyscale images with an alpha channel
  • Png now reads mono images as mono.
  • make install (via libtool) didn't work if "." weren't in the PATH.
• 30-Oct-2002
  • GenerateFamily() scripting command had a number of problems.
• 26-Oct-2002
  • The "Apple" checkbox was sometimes (incorrectly) invisible in the Generate Font/Family dialogs
  • Minor improvements to the "Insert Character" dialog (to match umap)
  • Minor tweak to text fields
  • More configuration changes
• 25-Oct-2002
  • Problems when reading an sfd with references to cleaned up characters.
  • Removed code from ttf generation that check for points of inflection (but fixed a couple of bugs in it before doing that).
  • Change the way scripting GenerateFamily behaves. One person has already been unable to figure out how to use it, and I agree that it didn't make much sense the way I had it.
• 24-Oct-2002
  • Change the default configuration process to use dynamic libaries
    • Added "configure.static" which will configure for static libaries (On Mac OS/X the compiler gets internal compiler errors if asked to generate pic code)
  • Enable reading a straight ttf file out of a .bin or .hqx wrapper (ie. doesn't need to be in a resource fork)
  • The Kern Pairs dlg did not show Hebrew/Arabic right to left. Fixed
  • Added a popup menu to Kern Pairs which will allow you to reset a kern pair
  • Added a way to configure the font family used for info messages in the outline and bitmap character views.
  • New man pages which use the macros supported by Sun. Thanks to R. P. Channing Rodgers.
• 22-Oct-2002
  • Initial version of Mac Family (FOND) generation
  • Fix old sfds so that their bad pngs will become bitmaps
  • optimize the fix to draw magnified.
• 21-Oct-2002
  • Verbose addition to scripting broke loops
  • Various changes to make utf8/ucs2 selections work properly with other programs
  • Hebrew/Arabic drag and drop could cause crashes
  • The Increment right side bearing command (Metrics->Set RBearing) set the width to the value rather than incrementing by that value.
  • Fixes for the local encoding being utf8
  • Imported bitmap pngs still were seen as indexed
  • Problems drawing magnified images when the scale factor was 100%
• 20-Oct-2002
  • Updated French and Japanese UI translations
  • Use LC_CTYPE (rather than LC_MESSAGES) to determine the current local encoding
  • If the width line were selected then FontForge would sometimes snap cursor movements to an arbitrary x-value (often the axis)
  • KANOU provided a patch which fixes some remove overlap problems.
  • I had broken the generation of pure bitmap sfnts from scripts (7-Sept)
  • Added a verbose mode to scripts (prints script to stdout as it is executed).
• 18-Oct-2002
  • Tweaked metrics view so that null kerning now displays as an empty string rather than "0"
  • If a font contained a kerning pair to a character that was omitted when saving the sfd file, then loading that font would leave it in a bad state and eventually cause a crash. Fixed on both ends (don't generate bad sfds any more, and can deal with them properly when found)
  • Improved the "Insert Character" dialog a bit.
• 17-Oct-2002
  • More work on Ik format files
  • Redid UI for Auto Width/Kern to make it easier for cyrillic/greek users to kern their alphabets.
  • Changed so that CJK encodings will attempt to map directly to a CJK font (rather than going through unicode) when displaying in the fontview.
  • Added a command to join paths
  • KANOU points out that support for mimetype cut/paste (11-Sept) broke cut/paste in scripting.
  • KANOU supplied a patch for the knife tool.
  • KANOU points out that IsVisible didn't work properly if passed a NULL point (it should never be passed a NULL pointer, but if it is no reason not to work correctly).
  • KANOU points out that the UniqueID for ttf files used a bad date.
  • Had the wrong MS language ID for hungarian
  • Some of Luc Devroye's fonts contain a "div" in the FontMatrix. ATM doesn't support this, but it is easy enough to accept. Still don't support a full expression there, just a div.
  • If we read in a gsf font and then generated postscript from that, we weren't able to read in the result.
  • Added an empty glyph-encoding sub-table to mac FONDs
  • Save multiple postscript did not save the first sub-font
  • The second count in BeginChars could be wrong in an sfd file (FontForge doesn't care, but I suppose some program might).
  • Add support for local encoding being UTF-8. Prefs now has an option to try and figure out the encoding from the locale
  • Display filenames in the local encoding.
  • There was an erroneous test in simplify which caused FontForge to complain about an internal error when in truth all was well.
  • There was a bad test in dumpbdf
  • Export will now allow the user to generate png files (if s/he's got the png libraries)
  • It is possible to invoke Export from a script
• 11-Sept-2002
  • Added a menu in textfields (right mouse button)
  • Added support for STRING_UTF8 selections
  • Added support for "image/png", "image/bmp", "image/eps", "text/plain;charset=ISO-10646-UCS-2", "text/plain;charset=UTF-8"
    • (two image format)
    • (one format for reading spline data)
    • (two more ways of expressing unicode text data)
• 10-Sept-2002
  • Added support for UTF8 in display/print dlg imports.
• 9-Sept-2002
  • Added support for loading Ikarus .IK font files
  • Fixed a divide by zero problem when generating ttfs
  • If you type into the Kern Pairs dlg, it will now scroll to the pairs associated with the character you typed.
• 8-Sept-2002
  • Display crashed on systems without freetype
  • my xfig reader didn't handle arcs
  • Merge fonts and Interpolate fonts didn't merge/interpolate kerning pairs
• 7-Sept-2002
  • The ruler tool will give the length of a spline if applied between the spline's endpoints
  • Made FontForge a little smarter about when to change the xuid field.
  • Added a dlg to display fonts using freetype.
  • Added support for the obsolete _XFREE86_GLYPH_RANGES property and then turned it off by default.
  • Improvements to type2 output.
  • Changed greek accents to use 0x384 rather than 0x30d when substituting for 0x301
  • I've found another case where Apple and MS/Adobe disagree on the format of ttf/otf files. So there's now a check box in the Generate dlg which allows you to select between the two defns. I've removed the two different kinds of bitmaps for ttf files, these are now controlled by this new checkbox too.
  • FontForge would sometimes in otf files generate curves which should have been done as lines.
  • macbinary resource output broke when I fixed straight resource output the other day.
  • When a grid line was a line but expressed with control points, then fontforge would get very confused about how to snap to it.
  • Fixed a bug in the generation of ttf/otf fonts with big5/sjis/wansung encodings.
  • The ligature dlg had a bad habit of crashing.
  • On the Mac, Obliqued fonts are stored in a weird postscript format. Try to support them
• 2-Sept-2002
  • When in no-x mode, fontforge did not notice any arguments (except -script). It will now notice -version, -usage and -help.
  • The PS Names check box would be visible when generating otf fonts. It should not have been.
  • If a font contained multiple encodings for the same character, then generating an otf font from it would crash fontforge.
  • There is now a proper french translation of the ui (Merci Pierre Hanser!)
  • Under cygwin, FontForge would fail to find internet explorer/netscape (because not in user's path) and if it did find them would try to invoke them with a cygwin style name (ie. /usr/local/share/doc/fontforge/overview.html rather than c:\cygwin\usr/local/share/doc/fontforge/overview.html )
  • Added a preference item to allow you to control whether a mapping file from glyph ids to character names is produced along with ttf and otf fonts.
• 1-Sept-2002
  • Added an #ifdef to an include file to make it work when built without X11
• 31-August-2002
  • Making a font be a cid-keyed font might screw up all other fontviews and lead to crashes.
  • type1 cid keyed fonts got the wrong fontmatrix if the em-size was something other than 1000. Doesn't seem to make a difference though, we still get bad output for emsize!=1000
  • got maxp.numGlyphs wrong for otf cid-keyed fonts which lacked a .notdef character
  • Added a check to make sure fontnames are less than 63 characters. Needed for ATM.
  • format 2 Coverage tables (used by GPOS and GSUB tables (kerning & ligatures for otf)) were being generated incorrectly. This meant that windows would refuse to read my otf files which had complex kerning. (presumably complex ligs, but I had no test cases). The mac printer driver has also had problems with my otf files. This may be why.
  • Added a popup to the kern pair dlg to tell user how many kern pairs there are, and how many characters start kern pairs.
• 29-August-2002
  • Merge and Elide were backwards
    • And merge (simplify) had problems when a desired spline at the same slope at the start and end (or something like that).
  • Added a field to Review Hints showing the total number of hints, flag for h/vstem3, ability to move points associated with hint when moving hint. also added a hot key for it.
  • Added a Find Problem to detect things which are almost but not quite h/vstem3
  • AutoHint wasn't finding all the ghost hints it should have.
  • Put in (what I hope is) a better way of detecting extraneous hints (the hypen case. At first sight the hstem and vstem appear equally valid) remove hints which are wider than they are long.
  • More fixes from KANOU for cid fonts
  • The scale to em checkbox only (sort of) worked if the emsize increased. Fix from KANOU
  • Autohint could go into an infinite loop on characters containing a 0-length spline.
  • Problem with cff name table fix of 27-Aug
  • bdf fonts were loaded into background by default
  • Scripting problem loading bitmap fonts into background.
• 27-August-2002
  • Further attempts to fix the fact that FontForge in some ways does not believe that monospaced ttf files are monospaced.
  • Problems with endif
  • Redid the Find Problems dialog by adding tabs so that related problems are grouped together.
  • Added a problem which will report advance widths which do not match some standard value.
  • CID keyed generation was broken again by the (failed) attempt to make things work with FontLab (18-Aug)
  • cff name table could be off by 1 if there were a real notdef character
  • Added ability to display blue zones in outline view.
• 21-August-2002
  • Broke cid-keyed otf generation on 27 July when I enabled the [Stop] button. Fix by KANOU.
  • Bringing up progress indicator early (20 August) broke scripting without X11.
  • Added ability to load a font protected by fstype=2 to a scripting (extra argument to open).
  • Enhanced print abilities.
  • Added ability to Print to scripting.
  • Attempt to fix the fact that windows doesn't recognize my monospaced ttf fonts as monospace.
  • FontForge could be confused by strange values of Ascent in the hhea table
  • Added a Round to Int option to the transform dialogs
  • Added ability to change the em-size (and rescale the font) to fontinfo
• 20-August-2002
  • Under KDE my tools palettes are made too wide and without borders. That's ok until they get reparented (docked). Added code in the docking routines to fix these problems.
  • May have fixed a race condition problem due to reparented windows under freebsd.
  • TeX sfd files are a little more complicated than I had originally thought, the parser is now more general.
  • FontForge sometimes thought some X fonts had iso8859-1 encoding when they did not.
  • cleaned up some problems with encodings
  • Added support for TeX (actually MetaFont) gf files
  • RLE background images sometimes showed up inverted (with background color as foreground & vice versa)
  • The Fit in window command should now take docked palettes into consideration.
  • Added a Japanese translation of the UI, courtesy of KANOU Hiroki.
  • Added the ability to run mf on a .mf font, generate a gf file, autotrace the gf file, and leave you with a font based on an mf file.
  • Menus now do pointer grabs.
  • Fixed a bug in generating PostScript resource file fonts on Mac OS/X
• 18-August-2002
  • Added a dlg to show all kern pairs (and another to show all ligatures)
  • Added a "Find Problems" item to check for missing bitmaps.
  • Put in some warnings to the ttf parser for bad cmap and glyf tables
  • FontForge was still having trouble about loading space bitmaps (that had been generated with freetype).
  • FontForge did not set the flag indicating manual hinting after changes by Review Hints or Create Hint. Should be set now.
  • Various small tweaks to the pfa/pfb output because FontLab doesn't like what we currently do. Probably none will make much difference.
  • The Print command would attempt to print characters outside the encoding of type1 fonts. This did not work well.
  • The autowidth command got very confused when given a sans-serif script to autowidth when the Latin characters in the font were serifed.
  • The Print sample text is now ordered so that the language specified in the LANG environment variable comes first in the list.
  • Added $version
  • FontForge will now read scripts from stdin if you follow the "-script" argument with "-".
  • Valek points out the using CharInfo to change a character's color did not change it until the fontview was refreshed
  • Expand Stroke had problems when intersections bent the wrong way.
• 9-August-2002
  • I've decided to encourage use of the mailing lists for reporting bugs. A feeble attempt to cut down on the spam I get by removing my e-mail address from the site.
  • Bug in the dlg asking for the resolution of a bdf font. The second (and subsequent) time(s) it came up it would have a good chance of never going away again, and potentially crashing FontForge into the bargain. (number of other bugs here too)
  • If the postscript "FontName" contained a hyphen then any generated BDF fonts would have invalid X fontnames (they hyphen would be included which would throw the name parsing off by one hyphen).
  • Greg Ford points out that generating a bdf font and then loading it back in will often change the fontname. This is because the fontname is not actually stored in bdf files. I've added a FONT_NAME property to the bdf fonts I generate.
  • FontForge would sometimes forget about the first bitmap character in each bitmap font stored in an sfd file. Please check any sfd files to see if they are missing bitmaps of space, exclamation point, etc. Probably introduced 11-June.
• 8-Aug-2002
  • After yesterday's fix textfields ignored backspaces
  • More work on the metrics lines. There's now a dlg to control which ones get drawn.
  • FontInfo would sometimes crash after adding ttf names.
  • Added minimal ability to dock palettes into their respective views.
  • Added a way of looking at the fontview where there are now holes in the font (ie. a compacted view where every cell contains a glyph). This is just another encoding, but people have been asking for it and it is slightly easier to get to than normal reencoding of a font. Note: It has no defined relationship to the glyph array in a ttf file.
• 7-Aug-2002
  • The 21-July autowidth/kern change broke autowidth/kern scripting.
  • And the serif detection failed if the character I did not reach the baseline.
  • FontForge read RGB png files incorrectly.
  • FontForge used to bring up most dialogs underneath the cursor. Now it only does that if the focus is set to FocusFollowsMouse, and otherwise centers dlgs.
  • FontForge now will draw metrics lines in the fontview if you want it to.
  • FontForge remembers what outline character views were open when a font was saved (as an sfd file) and will automagically reopen them when you next load that font.
  • Added View->Find In Font View command to the outline character view. Will scroll the fontview to display the current character.
  • Textfields within FontForge ignored the numeric keypad
• 4-Aug-2002
  • Yesterday's resize fix didn't quite work
  • Try to improve conversion of cubic to quadratic spline (actually, we make the conversion worse, but it probably helps rasterizers not to have points quite so close together).
  • Some more work on the TTF style name
• 3-Aug-2002
  • Failed to read ttf fontnames properly for some dfonts (probably some ttc files as well)
  • Didn't position references by point matching properly (when loading from ttf)
  • removed some debugging code inadvertently left in after the rle fix.
  • Added support for X Input Methods (to do CJK character input properly)
    This only works if X_HAVE_UTF8_STRING is defined -- an XFree 4.0.2 extension (ie. not on solaris)
  • Added a possible fix to a resizing problem for some window managers.
  • Added CIDFlattenByCMap function to scripting.
• 30-July-2002
  • Fixed a couple of problems found by valgrind
  • Applied a patch from KANOU Hiroki. PasteInto did not work.
  • Added run length encoding compression to sfd files (for background images). Bitmap image sizes are much smaller now.
  • FontForge failed to recognize that a PNG black and white image was monochrome.
  • Now FontForge will immediately put up the progress indicator when loading the first font (so the user has something to look at)
  • A little more work on generate multiple.
  • Removed a confusing message about autohinting a font for which no autohinting was actually done (it was just checking to see if it needed to).
  • Oops. Windows won't accept a ttf file if there is no name table entry with the same platform/specific entries as those in the cmap.
  • If a character contained a reference to a blank character that FontForge decided wasn't worth outputting, then the truetype composite for it would not work.
  • The standard format for the ttf "version" name is different from the equivalent postscript name. Added code to interconvert between the two rather than expecting them to be the same.
  • Added a gasp table for when there was no hinting.
• 28-July-2002
  • Pressing the [Filter] button on a blank field will show the current filter text (holding the mouse over the [Filter] button will cause a popup to appear with the same info).
  • If there were know selection the pressing an arrow key would presume that the selection was at encoding -1, which caused strange behavior. Now we will pretend that the selection is visible and halfway down the screen.
  • Oops. CMap files allow for multiple encodings for the same glyph. Now prepared to put in refs.
  • Oops. Generate multiple had several problems (couldn't handle encodings above 0x80000000, was surprised by empty font numbers, etc.)
  • If a family name contained a space then only the first word in it would be read out of an sfd file
• 27-July-2002
  • Oops. I was calling TeX sfd files Sub Font Directories and their proper name was Sub Font Definition.
  • Enabled the [Stop] button in the progress dlg when saving a font.
  • FontForge would go into an infinite loop when reading one of its sfd files when it was expecting a TeX sfd file. Fixed, and put in guards so that won't happen anyway.
  • Added two new menu commands "Flatten to CMap" and "Convert By CMap". These will use one of Adobe's CMap files to guide the Flattening/Converting process (rather than one of my cidmap files)
  • Added support for some really large encodings if they have almost equally large gaps in them (use adobe's "coderange" data to manage a sparce array).
  • If we could not find any style modifiers in a fontname, then we used to claim that the style was "Regular". I think it is better if we claim that it is /Width (ie. whatever is in the FontInfo /Width entry) and if that isn't set then return "Regular")
• 26-July-2002
  • Added support for Werner's (TeX's) Sub Font Definition files which provide a way of splitting up a 2 byte encoding into lots of little pfb fonts with 256 characters in each.
  • Normally in a ttf font, FontForge generates a 'post' table with the names of all the characters in it. I've just added a way to make it generate a short 'post' table with no character names.
  • FontForge generates bad truetype hints. There is now a way to turn these off in the Generate Font dlg.
• 24-July-2002
  • Code for finding the style still not working as it should.
  • Oops, the postscript parser was over-zealous about generating warnings.
  • Oops, the code which interprets results from autotrace was over-zealous on getting rid of contours in small images.

    ---

  • SetUnicodeValue did not behave as it should have.
  • Made acorn2sfd a little sturdier.

    ---

  • Bitmap copies didn't work from the font view (probably broke when I added anti-aliased support)
• 23-July-2002
  • Crash bug introduced yesterday by the duplicate name work.
  • Initial version of acorn2sfd -- a program to convert Acorn RISC/OS font directories into sfd files.
  • If a font is loaded from a ttf file, then it will have both postscript and ttf versions of certain names stored (family, copyright, version, etc.). The postscript names are obviously displayed, but changing them would not change the ttf versions. So generating a ttf font and then looking at it would still show the old setting of the name. The two should be better synchronized now.
• 22-July-2002
  • Introduced a crash yesterday if a postscript font contained an name in an encoding vector which was not defined in the font.

    ---

  • More work on duplicate names. I no longer append a ".<number>" to the end of duplicate entries and when generating fonts I remove them and generate an encoding with has multiple references to the base glyph. Which is what I should have been doing.
  • Small improvements to the style name detector.
  • If the Width field was a null string, then reading it from an sfd file caused us to thing the next keyword was the width.
  • Fixed a bug in the display of menus.
• 21-July-2002
  • Various small bugs in build accented characters
  • On the mac the filenames for mac resource fonts are different from same fonts elsewhere (elsewhere they have an additional ".bin"). Made either extension work from a script file (so they can work cross platform).
  • Expand Stroke had problems with curved splines where the slope at one of the end-points was 0/0. Since this will happen by default when joining a corner point to a curved point this should have been a frequent occurrence.
  • Bug in generating 'post' table if glyph 0 existed but was not .notdef.
  • Postscript fonts with one glyph going to multiple encodings cause problems.
  • Now both PS and TTF load will have consistent names in the above case.
  • My presumption of how symbol encodings were stored under MicroSoft was wrong. It is not a one byte encoding, it is a map into the private/corporate use area of unicode. 0xf000-0xf0ff. Changed to deal with that.
  • Loading a font with mac roman encoding often generated a font of 65536 characters (rather than 256+a few extras). Fixed.
  • New version of autowidth/kern. I don't think it kerns serifed fonts well, might do ok for sans.
  • Ability to provide autokern with a list of potential kern pairs and have it just check them.

    ---

  • Scripting support for kernpair files
  • AutoWidth/Kern now does a little better on serifed fonts
• 15-July-2002
  • Correct Direction should now only set the changed bit when it actually changes a character.
  • Correct Direction will now warn about flipped references, and offer you the chance to unlink them (after which it will be able to correct their direction)
  • Added ability to scripting to check if a character has changed, and to get a font's filename.
  • Cleaned up build accented chars a bit.
  • Another attempt to keep X from crashing on disappearing progress dlgs.
  • Maximizing the fontview could make it bigger than the screen so that the scroll bar was unusable.
• 14-July-2002
  • Valek points out that bitmaps generated by freetype were one pixel too high.
  • Changed the get info dlg for references so that the user could change the transformation matrix.
  • Font Info->PS Private got the wrong value when guessing the vertical stem snaps
  • FontForge did not find ghost hints for things at the numeric height.
  • FontForge got non-integer values for some hints where it should have found integer values.
  • Bug when auto-hinting references fixed.
  • FontForge would find hints for things that don't need to be hinted (finding a vertical stem for the top of "E" for instance. There is a vertical stem there, but it is deemed irrelevant). It now finds fewer of these hints. I hope it still finds all the legitimate ones.
• 13-July-2002
  • I think I've finally worked out Apple's scaled offset composite glyphs (and they do not behave as Apple documents)
  • FontForge was not setting the maxCompositePoints and maxCompositeContours fields of the maxp table of truetype fonts. This meant that in some cases rasterizers would crash when trying to deal with characters containing many components.
  • The translation tables between unicode and johab were wrong.
  • FontForge failed to generate ttf fonts with big5/sjis/wansung cmaps.
  • Changed so that reading sjis/wansung encoded ttf files would use a sjis/wansung encoding in fontforge (rather than the 94x94 encoding that no one but me seems to understand).
• 9-July-2002
  • Bug in generating ttf fonts with embedded bitmaps.
  • Open Font dlg will now try again if it can't load a valid font.
  • Started