diff options
Diffstat (limited to 'htdocs/changelog.html')
| -rw-r--r-- | htdocs/changelog.html | 472 |
1 files changed, 470 insertions, 2 deletions
diff --git a/htdocs/changelog.html b/htdocs/changelog.html index ff400cd3..e394c3ea 100644 --- a/htdocs/changelog.html +++ b/htdocs/changelog.html @@ -2,7 +2,7 @@ <HEAD> <!-- Created with AOLpress/2.0 --> <!-- AP: Created on: 27-Aug-2002 --> - <!-- AP: Last modified: 29-Apr-2010 --> + <!-- AP: Last modified: 20-Feb-2011 --> <TITLE>Change log for FontForge</TITLE> <LINK REL="icon" href="ffanvil16.png"> <LINK REL="stylesheet" TYPE="text/css" HREF="FontForge.css"> @@ -47,11 +47,479 @@ </BLOCKQUOTE> <UL> <LI> + 20-Feb-2011 + <UL> + <LI> + Did a lot of work to improve the accuracy in remove overlap. Improved accuracy + means that it is less likely to do the wrong thing. Problems happen when + there are points or intersections very close to each other. (So quadratic + glyphs are more likely to have problems than cubics simply because they have + more points, on average). + <LI> + Fix various error messages. + <LI> + Remove some obsolete documentation. + <LI> + Technical fixes to stroking code. + <LI> + Add a miterlimit to stroking code. + <LI> + FontForge was using the wrong MIME type for svg files. W3C has changed it + and it's now "image/svg+xml" not "image/svg-xml" or "image/svg". + <LI> + Since Inkscape images tend to be at unexpected y positions, force the view + to rescale/recenter itself if the result after a paste/import is outside + the current window. (Inkscape tends to put things at the top of a page, but + the origin is at the bottom, and FF looks at the origin. + <LI> + A line joint of 180 degrees (which shouldn't have happened in stroking, but + did) caused an infinite loop. + <LI> + Typo in add extrema code (again). + <LI> + Patch from Khaled Hosny: sets correct version of OS/2 tf table. + <LI> + Patch from Khaled Hosny: instead of dumping all OtfFeatName's into a single + line, dump each on a line of its own, this make VCS diffs more readable. + <LI> + Commas in the list of bitmap strikes could be recognized as a decimal separators + when generating fonts with bitmap strikes and this caused broken fonts to + be generated. + <LI> + In various places FF was supposed to temporary set LC_NUMERIC to `C', but + the original locale was actually never restored. + <LI> + Add a python reverseDirection() method to Layer. + <LI> + Make the head_optimized_for_cleartype flag accessible via Python interface. + <LI> + Patch by Khaled Hosny to allow setting default font file base name from python. + <LI> + Patch by Khaled Hosny: adds <font>.texparameters member to access TeX + font parameters from Python, readonly for now. The returned values can be + different from what is present in FF UI due to different rounding methods. + <LI> + Patch from Khaled Hosny: python's addAnchorPoint() function does now check + if a similar anchor point exists or not, and behaves accordingly. + <LI> + When we open an sfd file it can contain flags asking us to open charviews + for some glyphs. But on the mac, if we open more than one charview X11 (usually) + crashes. So, on the mac, put in a hook that only opens one. + <LI> + valgrind found a place where I was looking at freed memory. + <LI> + In the metrics view list index was not properly updated when user was manually + selecting an item from the word list. So wromg items were selected if (s)he + then used up/down arrow keys to navigate the list. + <LI> + Matthias Kilian spotted a buffer overflow in svg processing. + <LI> + The fix for LastResort.ttf had an undesired effect: each time an sfd file + was opened all AltUni entries were duplicated. + <LI> + Fixed validation crash for references to empty glyphs (Fedora bug 536920); + patch by Paul Flo Williams + <LI> + Patch by Paul Flo Williams to fix a crash when resizing toolbox while editing + bitmap glyphs. + <LI> + Try to get compilation working with multiple versions of libpng on the system. + <LI> + Khaled Hosny found that deleting the last anchor in a glyph causes a crash. + <LI> + Khaled Hosny fixed a crash in multiline text editing widget. + <LI> + Patch by Paul Flo Williams to fix a crash when attempting to edit glyphs + in Unicode Plane 16. + <LI> + Patch by Werner Lemberg to make FF compile with Python 2.7. + <LI> + Patch by Khaled Hosny to prevent crash when adding certain lookup types. + <LI> + If one try to compile with -Wall -Wextra, lots of warnings show up. The first + to fix are nested comments. As a result, a value that got, commented out + in gxdrawtxt.c, probably by accident, is now active. + <LI> + Notice from Andrey Panov: FF did not compile with gcc 4.4.4 + <LI> + Khaled Hosny made gtk stub compile again. + <LI> + Some typos and new comments in autoinstr code. + <LI> + Khaled Hosny provided an option in preferences/edit to turn off automatic + openning of autokern window for any new horizontal kerning subtable. + <LI> + Patch by Khaled Hosny to fix wrong type conversion when assigning GASP values + from python script. Spotted by Adam Twardoch. + <LI> + Khaled Hosny noticed that deleting anchors from Anchor Info dialog didn't + mark glyphs affected as changed. + <LI> + Patch by Khaled Hosny, who noticed that OpenType glyph positioning is applied + from left to right even for right to left scripts. As he writes: + <BLOCKQUOTE> + 'This patch fixes the kerning pairs dialog to render kerning as close to + what OpenType engines does, also makes sure FF activates by default and adjusts + the correct fields. There are other places in the code that need to be checked + too, but I think the code affected by this patch is self contained.' + </BLOCKQUOTE> + <LI> + Fixed a crash when FontForge was given NULL glyph name in python script. + <LI> + Fixed misinterpretation of '+-' in UTF-7 strings read from sfd files. Spotted + by Paul Flo Williams. + <LI> + Fixed misinterpretation of '//' at end of lines read from sfd files. Spotted + by Paul Flo Williams. + <LI> + The statemachine table uses a matrix gadget, but some of the code was still + expecting a list gadget, causing a crash. The patch, by Keith Stribley, also + fixes a couple of typos and round tripping of the glyphnames + <LI> + Patch by Keith Stribley, to fix crashing lookup metadata that don't have + scripts specified - he points that AAT doesn't need them. + <LI> + Patch by Keith Stribley, to fix corrupting of single substitution lookups + used in the state table after generating are corrupted. + <LI> + Patch by Keith Stribley, to fix export of Mac features, though it may not + cover every case. + <LI> + Patch by Khaled Hosny, to update Unicode support to Unicode 6.0.0 + <LI> + One point in code forgot to check if unicode number is in range before testing + for alternatives. + <LI> + Removed two infinite loops in overlap code. Turned out to be a floating point + accurracy issue. + <LI> + Zoom by mouse scroll scales around mouse pointer now. + <LI> + Ctrl+A in background layer (in charview window) picked up anchor points from + foreground layers, too. + <LI> + Fontforge forgot to categorize points after importing a svg font. + <LI> + Patch by Philipp Stephani, to fix wrong interpretation of most top accent + data for newer versions of Cambria Math. + <LI> + Patch by Louis Simard: fixes multiple stack based buffer overflows in BDF + processing routine. Ref: https://bugzilla.redhat.com/show_bug.cgi?id=659359 + <LI> + Patch by Barry Schwartz to fix error in handling of glyph name ranges in + feature files, reported by Matthew Skala: + <LI> + Fixed crashes on pasting splines into guide layer in the glyph view window, + reported by Andrey Panov. + <LI> + Remove one source of rounding errors in solving cubics, clean up function + calls to match new calling conventions. + <LI> + Matthias Kilian notices that previous patch to fvimportbdf.c made unterminated + strings. + <LI> + Khaled Hosny spots a paragraph in the docs that's no longer valid. + <LI> + Cleanup the fontforge desktop file. + <LI> + Freehand tool was very broken. + <LI> + When doing remove overlap on quadratic beziers (truetype) ff converted them + to cubics and then did the ro. This introduced slight rounding errors which + gave ro canniptions. Just as easy to make remove overlap work on quadratics + too. + <LI> + See if we can improve conversion from quadratic beziers to cubic. There are + rounding errors involved in finding the "cubic" control points which mean + that the cubic term of the spline will be non-zero when it should be zero. + See if we can find better approximations to the control points, and be prepared + to force a small third order term to 0. + <LI> + "Make arc" didn't work as expected for connected points (it actually produced + a straight line). + <LI> + The previous patch for matrixedit was incomplete: although the function indicator + box was always drawn at the visible area of the control, it actually didn't + work. Moreover, the box was being duplicated multiple times during horizontal + scrolling. So this is one more attempt to make the whole GUI for editing + OT glyph classes functional again. Note that now the special handling is + applied only to the last column of a matrix edit control: if this column + contains text fields with function indicators, then those indicators are + always drawn at the top of the field by a such way that they should always + be visible. + <LI> + Daniel Dumitriu has provided Romanian translations for style names. + <LI> + Daniel Johnson sent me a print sample of Mayan in the K'iche' dialect used + in Guatemala from the prolog to the Popol Wuj. + <LI> + Expand stroke didn't work if the intial spline in a contour had a length + of 0. + <LI> + Khaled points out that if the subtable dlg for a single substitution is + [Populated] then hovering the mouse cursor over an empty cell will cause + a crash. (the thing that pops up an image of the sustitution failed here). + <LI> + Legacy scripting did not support importing pdf files. + <LI> + There used to be two preference item controling autohinting. That does not + seem a good idea, so coelesce them into one preference item. + <LI> + _strtod could munch the stack if given a long string. This caused crashes, + etc. + <LI> + Fixed a number of problems with the mm dlg: + <OL> + <LI> + of the design vector were bigger than 60 characters then ff would crash (happened + if there were more than 2 axes) + <LI> + default values for the Slant access were backwards + <LI> + Convert-Design-Vector functions were wrong for 3 & 4 axis designs. + </OL> + <LI> + The freehand dlg did not retain the previous shape of a polygonal nib. It + always used the default shape each time it was invoked. + <LI> + More problems with zero length splines in expand stroke. Biggest problem + was that the routine which was suppposed to remove them didn't (at least + in some cases). + <LI> + Add a link to the panose docs from the panose pane of fontinfo. + <LI> + Add strings and support for other latin panose families. + <LI> + Patch by Werner Lemberg to improve rasterization accuracy in TTF debugger. + <LI> + Trying to create a new kerning subtable from font info would often crash. + <LI> + Brian Zick wanted a preference item so that if one end of a spline were moved + and the other were not then the control points would be interpolated (rather + than being moved by the same amount as the adjacent end point). + <P> + Note that this will have problems if the moving end point ever ends up on + the same horizontal/vertical line as the stationary end point. Then the control + point may vanish and never reappear. + <LI> + Add a bit more info to the version string (whether compiled for doubles, + etc.) + <LI> + When modifying OT classes by inserting glyphs from a fontview-based window, + the add_char_to_name_list preference item was not respected. + <LI> + FF used to allow modifying the width (and vwidth) of a bitmap glyph only + in bitmap-only fonts. This seems to make no sense for TTF, where widths of + rasterized glyphs can be set independently for each PPEM. + <LI> + Use FT_LOAD_TARGET_MONO instead of FT_LOAD_MONOCHROME for spline rasterization + with freetype. This guarantees more accurate rendering of hinted fonts in + monochrome mode. + <LI> + If a font contains several sets of kerning classes (e. g. for different scripts), + and a particular glyph is present in more than one set, then attempting to + find a matching kerning value between this glyph and a subsequent glyph might + result in selecting the combination with the {Everything Else} class from + the first set being processed, even if there are better matches in other + sets. + <P> + Also FF would still complain that the "{Everythin Else}" string in the kern + class dialog represents a missing glyph. + <LI> + Two small corrections for metrics view: + <OL> + <LI> + Since the {Everything Else} class is now recognized by FF as a valid glyph + class, it was possible to occasionally modify a kerning value involving this + class in the metrics view. So when asking user if he is going to modify a + class-based kerning value or create a new kerning pair, inform him if one + of the glyphs from this pair has been actually matched by the {Everything + Else} class. + <LI> + Make it possible to easily deselect all glyphs in the metrics view by pressing + [Escape]. + </OL> + <LI> + Pasting bitmaps from fontview didn't work. + <LI> + Allow a user to enter values for the PS private dictionary in his/her own + locale but convert them to the "C"==PostScript locale when storing them. + That is a French user could enter [3,5 20] for BlueValues, but it needs to + be converted to [3.5 20] when put in the sfd file (or PS font file). + <LI> + Change the Histogram dialog to use boxes (cosmetic change, shouldn't matter + to English speakers, dlg won't have fields overrunning other fields in French). + <LI> + Redo the "PS Private" pane of the Font Info dialog so that it uses a matrix + edit rather than the old clumsy list with editing area kludge. + <LI> + Patch by Khaled Hosny to fix crash when importing svg with linked image. + <LI> + The scripting function "FontsInFile" did not support pdf files, and always + returned an empty array when asking if they contained fonts. + <LI> + Scripting didn't let anyone pick a subfont within a pdf file. (did not support: + filename(fontname) notation for pdf). + <LI> + Update cidmap to CNS-1.6. + <LI> + My cidmap files have always been a one to one mapping of cid to unicode. + But that's not what Adobe does. Many cids get mapped to several code points + each. So I've revamped the format to support this. The new map files need + the new code (obviously) so don't download one without the other. + <LI> + An earlier patch added support for horizontal scrolling (shift trackwheel) + in the outline glyph window. This builds on that and supports it in (I hope) + all windows. Also adds support for apple's wheel which can do both horizontal + and vertical motion. + <LI> + When ff calculated the size of a kerning subtable it was off by two bytes + per glyph in the coverage table. This meant that it would sometimes make + a table that was slightly too big to be supported and not realize it. I've + fixed the bug, and put in checks so that if there's yet another problem like + this, at list it will be reported to the user. + <LI> + It is now possible to differentiate stems by width (rather than by direction) + in the "Change glyph" dialog. This approach produces better results for glyphs + like "N" or "g", and thus seems more logical even for the Latin script, where + vertical stems are normally "thick" and horizontal ones are "thin". + <P> + This patch also affects some gdraw internals, since the new layout of the + "Change glyph" dialog makes it necessary to dynamically switch the captions + of some labels, and the text strings should fit to the space initially reserved + for those labels. So it is now possible to assign a list of strings to a + label, just like if it was a list button. + <LI> + Fix some issues with hv points. + <LI> + Do some internal cleanup. + <LI> + User doesn't like that ff does a "fit to window" ever time the user resizes + the window. So... let's try not doing that. + <LI> + Someone complained that when the fontview was maximized it would make itself + one cell smaller than it should have. Well that's screen size and window + manager dependant. It did allow 10 pixels of horizontal border (not knowing + what the WM would give it). That's probably too much. And even if a tiny + bit of the window is off the screen, it probably doesn't matter much. + <LI> + When ff was started with a .ufo fontdirectory on the command line, it would + fail to open it (and produce a garbage error message). I suspect the underlying + problem would cause other problems too, but haven't found any yet. + <LI> + Barry points out that FF incorrectly loaded an opentype font with an empty + bluevalues array (FF showed it as not having a bluevalues array at all, the + two are different). (pfb files with empty arrays worked as expected). + <LI> + Correct two "bugs" in UFO output. I believe that I was following the spec + in both cases, but since the reference implimentation does not follow the + spec I shall do what it does instead. + <OL> + <LI> + The spec gives a complex algorithm for converting a glyph name to a filename. + The proper behavior appears to be "add an underscore immediately after every + capital letter" and leave it at that. + <LI> + The spec says for OS2 fsType we should output bit numbers. It doesn't mean + the numbers of the bits (0..15) but rather the values (0..1). + </OL> + <LI> + Retain the UFO ascender and descender fields. + <LI> + Werner found a place where apple kerning output didn't work if glyphs which + weren't being output had kerning pairs associated with them. + <LI> + Richard Gitschlag and Khaled complain that the Glyph Info dlg says "Done" + rather than "Cancel". + <P> + There is a reason for this. The button does not do a cancel. It only cancels + changes to the current glyph, but the dlg can be applied to several glyphs + one after another, and any changes made to previous glyphs are committed + when the user changes glyphs. So the "Done" button only cancels the current + glyph. + <P> + However, I can change it so that it says "Cancel" until the user changes + glyphs. And in most cases the user is only interested in one glyph so this + will solve most of the problem... + <LI> + Richard Gitshlag suggests that the buttons in the "non-integral coordinates" + question of validation are confusing and suggests they be relabeled. + <LI> + More fixes needed where *printf expects a double and gets a long double if + --enabel-longdouble. + <LI> + According to the spec there are two different types of GPOS 'size' feature. + One specifies a design size and nothing else, the other specifies a design + size, a design range a style name and a magic number used to associate all + fonts in the same family with the same weight and style but with different + design sizes. + <P> + Fontforge was not very good about warning the user when s/he created something + which had some but not all of these attributes. (ie. a design range but no + style name). So add some error messages to the fontinfo dlg which makes these + checks. + <P> + FF would also not read in a 'size' feature unless it were in one of these + formats. So if the user created a 'size' feature with a design size, and + a design range but no style name, then FF would not complain, would happily + output it, but would not read it back in again. So be a little more lenient + in what we read in. + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + <LI> + </UL> + <LI> 13-Feb-2011 <UL> <LI> Switched from cvs to git. - </LI> </UL> <LI> 29-Apr-2010 |