| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
this, how can it possibly fail?).
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11175 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
build. This commit fixes it.
Revision 11146 changes the code to find the current op_array tables by
looking in the current instance. Unfortunately, during initialisation
of the postscript interpreter the instance is created and the init file is
read without the newly created instance being made 'current'.
The fix used here is to arrange that any newly created instance becomes
the current one automatically.
No changes shown in the localcluster regression, but this doesn't test
the language switch build. Seems to work OK on windows though.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11174 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
this commit reduces the reliance on mem_err_print.
Currently all calls to errwrite and errflush get the stdin/out/err details
from a global mem_err_print. For multithreaded gs this will result in output
going to the wrong place if any of them are redirected. This review splits
errwrite into 2 variants, errwrite and errwrite_nomem. The first takes an
explicit mem pointer. The second continues to assume a global one.
The plan is to make everything possible use the first. Currently debugging
calls and some 'stubborn' devices continue to call errwrite_nomem.
Local cluster testing shows no differences.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11173 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
introduced with making the font cache part of the library context,
revision 11165.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11172 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
(bug #690702).
Minor modifications were made to the original patch to make it work with the
current trunk.
Aaron Gifford's SHA-2 implementation is included (BSD licensed).
Libidn is added as an optional dependency to support unicode passwords. If it
is not installed, only ASCII passwords are supported.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11171 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
Bug #691274 "Missing or incorrect ToUnicode when using Identity ordering"
pdfwrite emits ToUnicode CMaps in PDF files which always use 2 bytes for the key
length. That is 0x21 is represented as <0021>, since these are hex strings there should
in fact be no difference between <21> and <0021>. It is clear from the ToUnicode
Mapping file tutorial that this is correct for mapping files.
However, a PDF file is not a CMap file. It appears that Acrobat does not like
beginbfrange data which is padded with zeros (it seems happy with codespacerange and
destination codes which are 0 padded though).
This patch outputs ToUnicode CMaps which are not padded with 0, ie 0x21 is emitted as
<0021>, for the srcCodeLo and SrcCodeHi parameters in a beginbfrange operation. The
destination codes remain zero-padded.
Technically this would be incorrect for a real CMap file, but we don't write those
out so this should not be an issue. The change is limited to ToUnicode CMaps and does
not affect any other kind of CMap.
I believe the change will affect the data emitted for the beginbfrange,
beginbfchar, begincidchar, beginnotdefchar, beginnotdefrange and begincidrange
operators. It does *not* affect the begincodespacerange operator; although Acrobat
Distiller writes this without zero padding tests show that Acrobat is happy reading
zero-padded code space ranges.
ToUnicode information is not used in rendering PDF files, so regression testing should
show no differences. Local testing of a limited number of files doesn't reveal any
problems, but this change has not been well tested.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11170 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
committed for bug 690695
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11169 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11168 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
instructions for legacy old systems being deprecated; bug 691264.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11167 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
between local cluster regression run results between arbitrary revisions.
Usage information given in file. Must be run on the cluster master machine
(i.e. casper).
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11166 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
global but part of the library context.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11165 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
Bug #691259 "HPGL graphic followed by raster graphic results in incorrect pdf"
The pdfwrite device is initialised in 'C', and then partly overridden by PostScript
initialisation. The problem is that non-PostScript input does not (cannot) execute the
PostScript initialisation, and so the defaults are left as those done in 'C'.
Not only do these defaults not match the documented defaults in ps2pdf.htm, they are
not particularly well selected. In this case the problem was caused by sending image
data at 150 dpi, but only a single scan line at a time. The default setting for
pdfwrite was to downsample image data to 72 dpi, this caused us to subsample (the
default downsampling) by a factor of more than 2, which led to our single scan line
becoming less than half a scan line, and so disappearing altogether.
This patch addresses this issue by altering a number of the defaults to more sensible
values. In particular we disable downsampling and enable image compression for gray
and colour images. This has the fortunate result of making most PDF output files
smaller than before, even though we are now storing more image data.
Expected Differences
This change does unfortunately result in a *lot* of changes in our testing. Over 500
of our standard test files now render differently, too many to list here.
I have looked at the output of these and I believe the changes to be progressions, and
extremely worthwhile ones. A very few of the files now containing higher resolution
images do render less well at low resolution (75 dpi). This is simply an artefact of
the way image data is rendered. The same files render better than before at higher
(300 or 600 dpi) resolutions. Also many of the new, higher resolution, files render
better than before even at 75 dpi.
So overall I believe this is a very useful improvement.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11164 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
incorrect version number but Acrobat renders them just fine. Bug 691273.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11162 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
overwritten later.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11158 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
fixes them.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11156 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
compiler warnings.
No expected differences.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11155 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
allow libctx's to be shared between instances, and we'll never want to do
that. It's unused at the moment anyway, so this doesn't impact anything.
No expected differences.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11154 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
check is
complied out and replaced with DO_NOTHING in release build so this has effect only
on debug builds. Fixes Bug 691271 .
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11151 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
No local cluster differences reported, or expected.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11150 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
incorrectly (bug #691014).
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11149 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
font descriptor. The latter may be shared by font resource dictionaries with
different encodings causing incorrect rendering. Bug 690714.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11148 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
globals and into i_ctx_p. This requires various knock on changes
throughout the code to enable them to be recovered from the context
(which may itself need to be recovered from a memory pointer).
Also, the garbage collection is altered slightly; rather than the
op_arrays being special cased (as special objects that have to have
their component parts registered as roots for the garbage collection) we
move them inline into i_ctx_t and amend the gc routines for that
structure. Thanks to Alex for his help with this.
No differences expected, or shown with local cluster testing.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11146 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
(Make it an enum instead).
No changes expected or shown by local cluster testing.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11145 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11143 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
yesterdays commit of such fixes.
No expected diffs.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11142 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
This would very rarely have been a problem in current code, but would be
more of an issue in multithreaded operation as running out of memory
while forking a new instance is more likely.
No expected differences, but the local cluster code seems to be claiming
differences for an unchanged tree at the moment, so if 3 or so non-pdfwrite
errors and 1800ish pdfwrite errors show up, it's not because of this.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11141 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11140 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11139 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11138 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
and depend on -Zm for the FAPI warnings or -Z1 for the Freetype warnings. Also address compiler warnings in fapi_ft.c.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11137 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
explicit casts.
No differences shown by local cluster testing (of pcl and gs only as xps is
currently broken).
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11136 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
removing all global variables. Thanks to James Cloos for the patch. Fixes bug #691241.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11135 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11133 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
If the path is given as /path/_rels/.rels then
it will parse the entire document. If the path points
to a FixedPage file (/path/Documents/1/Pages/1.fpage),
only that single page will be parsed.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11132 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11131 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
The special mode of the XPS interpreter that parses the
ZIP file incrementally as a stream has proven to
be useless. None of the files we see in the wild make
use of the features that are required for a streaming
XPS parser to behave well. As a result, this mode was
always buffering up data and using lots of memory.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11130 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
proper deallocation. Current allocation may be reset when a PS error occurs.
Bug 691167.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11126 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
its contents.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11125 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
action only occurs if there is a change from the current overprint state. This fixes a problem in one of the files associated with bug 691014. In the particular file, overprint compositor actions were getting placed into a pattern clist. The file in fact had no overprint settings. Ran regression test, no differences expected.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11124 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
Bug 691217.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11123 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
memory.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11122 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11120 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
fonts installed on underlying platform. This may be needed to ensure consistent
rendering on the platforms with different fonts, for instance,
during regression testing. Bug 691240.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11119 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
present and a pattern with no transparency. This means the PDF14 device is installed and more importantly that transparency state pushes and pops occur to keep the softmask in sync with the graphic state. This particular files pattern is installed as a clist, which happens to not have any transparency. The transparency state pushes and pops were getting placed into the pattern clist when the Q and q operations occurred in the pattern, which was causing problems since the pattern clist device had no pdf14 device installed to handle such commands.
The solution is to check, when a transparency compositor action is to occur, if the current device is a pattern accumulator or a pattern-clist device. If it is, then the pattern instance is checked to see if it has transparency. If it does not, then the transparency state chance is ignored. This all occurs on the clist writing phase in the gs_trans functions.
In addition, a fix was made so that the pdf14custom device sets up it's color information correctly. For subtractive devices that do not have cmyk process colorants the custompdf14 device is installed. This pdf14 device was obtaining its color_info from the target device. The pdf14 device must be continuous tone even if the target device is not.
Ran a regression test and saw no differences.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11117 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
dictionary overflow in some cases. Unlike earlier array, dictionary
representation of 'cmap' table doesn't need this. Bug 691253, customer 780.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11116 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
Freetype enabled is now the default for GS.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11114 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
Freetype enabled became the default for the normal GS build.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11113 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
object was not getting freed. The large mask buffer, which is a member variable of maskbuf was getting freed but not this 12 byte structure. No expected differences.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11112 a1074d23-0009-0410-80fe-cf8c14f379e6
|
|
composite font in PDF file. Bug 691253.
git-svn-id: http://svn.ghostscript.com/ghostscript/trunk@11110 a1074d23-0009-0410-80fe-cf8c14f379e6
|
