[gs-devel] Easy to fix bugs that need some discussion.

[Alex Cherepanov]

Colleagues,

There are a few bugs I'd like bring to the support meeting.
The cause of all these problems is well-known but the resolution
of the problems needs some decision that I cannot make
on my own.

Regards,
Alex Cherepanov


687475 font glyphs overlap
Ghostscript makes poor font substitution. We can make a better font
substitution if we associate font names with character widths and
basic font attributes: weight, italic, serif/sanserif, etc. The new
font substitution code will take 1..2 days to write.

687643 gs crashes on HP-UX 11.23 for Itanium
This is an old issue that has been discussed several times already
on the merits of known patches.
At 2004-12-09 18:00 Ralph posted a comment #9 that Raph has a fix
but the fix itself has never been posted. Perhaps, this is what SCOX
call a non-public contribution. How can we discover the patch?

687682 Old failures in Genoa tests
This bug is closed but not all issues identified by the test have been
resolved. The following test failures seems to be errors in Ghostscript.
068-01 : low precision of the current point
078-01 : The test wants initial strokeadjust to be false
094-01 : The test expects that <</DefineResource {exch pop}>>
          is a working definition. Surprisingly, it works on Distiller.
255-01 : The test requires (%stdout)(w)file to return a local object.
          On Ghostscript, it is global.
272-01 : (%stdout)(w)file should return a local object even in the
          global mode.
421-01 : The test has a mixed Type 3 and Type 42 font. Igor, check how
          does this work on Adobe.

687822 /limitcheck in --setcharwidth--
GS suffers from low precision and small range of the fixed type.
The effect is subtle and difficult to spot in prepress.
Many problems can be resolved by using 64-bit fixed numbers with
32,32 format. The fix is easy on 64-bit platforms but 32-bit platforms
convert between fixed and long all over the code.
Do we want 64-bit fixed type?

688037 Large file support in Ghostscript. In my opinion Ghostscript
should be able to read and write large files on the platforms that
provide the API. The platforms include GNU+Linux and Windows.

687292 Intel C warnings
Intel C found 1 real bug and a few minor issues. We also need to decide
how to print 64-bit numbers in a portable way.
Ralph, the bug is assigned to you. Please review the patch.

688385 pdf font error /undefined interval index
The patch is ready and awaits review. The patch was originally
developed for a low priority bug 688290 but become a high priority
customer problem after aging. Please review.

687297 New free font release needed
I can take over this project.

688268 implementation of Tn5044 operators
The remaining Tn5044 are coded and the patch is attached.
Dan, please review the patch.

687221  A tool for image comparison.
A cross-platform tool can be coded using wxWorks library.
Leonardo says that (1) there's something already coded in Python,
(2) we don't need a cross-platform tool. IMHO, Ghostscript is
free software and should not depend on Windows-only tools.

687498 Partial fix for not handled Acroform flag "NeedAppearances"
I've done it for a 3rd party. Please review.

688159 Problem decoding PDF with JPXDecode data
I can implement the approach, proposed by Ralph.

688432 Loss of bold in substituted font of pdf missing embedded font
So should I get the bold property from the font name and combine with
force-bold flag?

688412 Many new CMap's are available.
Many new new CMap resources are available. All files have higher or
equal version number than ours. All files are freely redistributable.

687986 Incorrect code in HEAD spdiff.c
Igor has closed the bug but the filter is still broken.
Should we remove the filter completely?

688464 Ghostscript displays PDF but gives out error messages
The font defines /SD and /ND in userdict. The patch puts userdict
higher on the dictionary stack and avoids the false warning.
Unfortunately, this disables font validation one /SD and /ND
are defined in userdict.

688469 Automatic cache size detection causes indeterminism.
On my GNU+Linux box, the value of ARCH_CACHE1_SIZE varies from 64K to 
512K from compilation to compilation. So the configuration program
should be fixed. The variation of  ARCH_CACHE1_SIZE also causes
rendering differences in some files, which is a separate bug.

688472 /ioerror in --token-- on page 39 of PDF file
The file interprets correctly when the streams are limited to the length
given as a Length attribute. Unfortunately, doing this every time is not
possible because the Length attribute can be incorrect. Perhaps, the
limit can be set only when the Length attribute is correct (i.e.
the stream has 'endstream' at this offset) and the first filter is
FlateDecode.