Age | Commit message (Collapse) | Author |
|
|
|
|
|
Add a cached color converter mechanism. Use this for rendering meshes
to speed repeated conversions.
This reduces a (release build to ppm at default resolution) run from
23.5s to 13.2 seconds.
|
|
In the existing code for meshes, we decompose the mesh down into
quads (or triangles) and then call a process routine to actually
do the work. This process routine typically maps each vertexes
position/color and plots it.
As each vertex is used several times by neighbouring patches, this
results in each vertex being processed several times. The fix in
this commit is therefore to break the processing into 'prepare' and
'process' phases. Each vertex is 'prepared' before being used in
the 'process' phase. This cuts the number of prepare operations in
half.
In testing, this reduced the time for a (release build, generating ppm
at default resolution) run from 33.4s to 23.5s.
|
|
When we meet a broken PDF file, we attempt to repair it. We do this by
reading tokens from the file and attempting to interpret them as a
normal PDF stream.
Unfortunately, if the file is corrupt enough so that we start to read
from the middle of a stream, and we happen to hit an '(' character,
we can go into string reading mode. We can then end up skipping over
vast swathes of file that we could otherwise repair.
We fix this here by using a new version of the pdf_lex function that
refuses to ever return a string. This means we may take more time
over skipping things than we did before, but are less likely to
skip stuff.
We also tweak other parts of the pdf repair logic here. If we hit a
badly formed piece of data, clear the num/gen we have stored so that
the next plausible piece we get does not get assigned to a random
object number.
|
|
Remove code that's not used any more as a result of the previous
fix, plus some code that was unused anyway.
|
|
The 0 null object is leaked if a document refers to 0 0 obj before
requiring a delayed reparation (seen e.g. with 3324.pdf.asan.3.2585).
|
|
Thanks to Simon for spotting the original problem. This is a slight
tweak on the patch he supplied.
|
|
Replace an explicit i = i by a comment in a for loop where i is
already at the correct starting value.
|
|
Make single-point strokes display by special casing them as
circles. Thanks for Michael Cadilhac for the suggestion.
|
|
Use round caps and joins so as to better match the result of drawing, and also
so that single dots display. Thanks to Michael Cadilhac for the suggestion.
|
|
In some cases freshly-created annotations could fail to appear because the HQ patch was
being left in place even when zoomed fully in, and when in that state, the patch was not
updated. The bug was usually hidden by an onLayout call being triggered with an out-
of-date patch, which causes the HQ patch to be removed. The bug is fixed by having
addHq remove the patch when fully zoomed out. Since now addHq may sometimes add
the patch and sometimes remove it, I've renamed it to updateHq.
Correctness of this fix has not been checked because I was unable to trigger the bad
behaviour on my test device.
|
|
Avoid unnecessary copies. Minimise calls to isbigendian.
|
|
This is required for e.g. 1980_-_compressed_inline_image.pdf and
Bug690300.pdf .
|
|
At https://code.google.com/p/sumatrapdf/issues/detail?id=2436 , there's
a document with an empty xref section which since recently causes a
repair to be triggered. Repairs then stop when pdf_repair_obj_stms fails
on an object which isn't even required for the document to render. Such
broken object streams should rather be ignored same as broken objects
are ignored in pdf_init_document.
|
|
Avoid recursion to avoid stack overflows.
|
|
The pattern repeat calculation should be done in pattern space, but
one of the arguments in the calculation was being taken from device
space. Fix this. Also only apply the bias in the case where the
bias would make it larger.
173 progressions.
|
|
Currently, if we spot a bad xref as we are reading a PDF in, we can
repair that PDF by doing a long exhaustive read of the file. This
reconstructs the information that was in the xref, and the file can
be opened (and later saved) as normal.
If we hit an object that is not in the expected place however, we
cannot trigger a repair at that point - so xrefs with duff offsets
in (within the bounds of the file) will never be repaired.
This commit solves that by triggering a repair (just once) whenever
we fail to parse an object in the expected place.
|
|
Empty Contents streams are not valid - they need a length at least.
The alternative approach would be to put /Length 0 and update it
later.
|
|
Thanks to Michael Cadilhac for spotting this.
|
|
Simple typo. Thanks to Alexander Monakov for spotting this.
|
|
Previously we were setting blendmode in the created form XObjects
transparency group definition. This didn't work as PDF readers don't
look for it there.
Now we set it in the calling stream's resources, and set it before
calling the group.
|
|
Thanks for Francesco Turco for spotting this.
|
|
Thanks to Makoto Fujiwara for spotting this.
|
|
It seems that (int)-98.5 = 98, not -99. Use floorf instead.
|
|
While scrolling, avoid some overheads to do with image scaling that
need updating only on a zoom-level change
Remove a pointless invalidate call.
Avoid calls to removeViewInLayout and removeAllViewsInLayout that
were being made in functions not called from onLayout
|
|
If we have a NULL page, don't attempt to pass events to it.
|
|
Thanks to Dale King for reporting this.
|
|
|
|
Unused field. Also tweak some comments for clarity.
|
|
|
|
|
|
|
|
issues with surface thumb rendering.
Also, make updates for changes made to font names and 3rd party file structures.
|
|
|
|
These were the last few fixes that went back from mupdf to gs.
This seems to solve bug 694362. Thanks to Shelly for this.
|
|
Thanks to Shelly for pointing out that this needed to be done. This
might solve bug 694362.
|
|
Only -I the config header directory if building the thirdparty library,
not if using the system library.
Fix bug 694808.
|
|
The OpenJPEG in gs is v2, with various patches for fixes. These are in
the process of being passed upstream. We now automatically pull the
openjpeg tree out of GhostPDL and put it in as one particular branch
in the thirdparty/openjpeg.git repo. Change to track this in MuPDF.
This is in keeping with what we have been doing with the jbig2dec
repo for a while now.
|
|
A poorly formed string can cause us to overrun the end of the buffer.
Now we check the end of the string at each stage to avoid this.
|
|
We were miscalculating the offsets into a sampled functions table,
causing us to overrun the end. Fixed here.
|
|
The changes to fz_render_glyph cause the scissor rectangle to no longer
match the transformation matrix which causes Type 3 glyphs to be
clipped at larger resolutions.
|
|
ft_file was removed in a2c945506ea2a2b58edbde84124094c6b4f69eac even
though it might still be needed by downstream consumers (such as
SumatraPDF) for allowing devices to load fonts again when a font has
been loaded by fz_new_font_from_file which doesn't maintain a buffer.
|
|
fz_new_font_from_buffer keeps the buffer for the font, so callers which
no longer need the data have to drop the buffer themselves explicitly.
|
|
The actual issue here is that a pixmap is dropped more times than
it should be due to an error in the rendering pipeline.
The problem arises because we fail to push a clip image mask, but
still pop the mask off the stack later. This puts us off by 1 in
the stack handling.
The simplest solution to this (that will be safe no matter what
mistakes are made by the caller too) is to add some simple tests
in the draw device to ensure we do not free too early.
|
|
I believe the implementation for revision 3 is wrong.
From pdf_reference17.pdf, step 5 of Algorithm 3.5 says:
5. Do the following 19 times: Take the output from the
previous invocation of the RC4 function and pass it
as input to a new invocation of the function; use an
encryption key generated by taking each byte of the
original encryption key (obtained in step 1) and
performing an XOR (exclusive or) operation between
that byte and the single-byte value of the iteration
counter (from 1 to 19).
"the original encryption key (obtained in step 1)" is pwbuf
(32 bytes) not key. Even if it was key, it wouldn't be n
bytes long, but only 16.
|
|
|
|
In case of an unknown function type, we free 'func'. Then we later
read func->type out of the block, and drop the block.
Simple solution is not to free the block initially and to let the
drop of the block do it for us.
|
|
|
|
|