Bug#796490: [texlive-latex-base] pdflatex does not create reproducible pdfs with multiple images
On Tue, Sep 01, 2015 at 08:23:07AM +0900, Norbert Preining wrote:
> > I can still reproduce it inside a sid schroot (uptodate) and than run
> > testscript.sh
>
> Yes, this is a different item. The first - fixed by updating - was
> about differences in the fine dict.
>
> The problem with images is not solved by now. I have no
> idea where this comes from.
>
> I checked pdftex.def which is responsible for image inclusions,
> but I don't see anything there. So it might be something in the
> actual source code
I think I found the cause of this bug. It's a bug in pdfTeX that results
in uninitialized memory being copied to the output file. I have reported
the bug upstream:
https://www.tug.org/pipermail/pdftex/2017-March/009100.html
The function write_png_rgb_alpha allocates twice as much memory as is
necessary for the smask buffer. The second half of the buffer is left
uninitialized and the whole buffer is copied to the output PDF file. It
only arises with PNG images that have an alpha channel. (I suppose
optipng removes the alpha channel when possible, which is why it made
the problem go away.)
I think the bug is in texk/web2c/pdftexdir/writepng.c, where a "/ 2"
should be "/ 4"; i.e., 1 in 4 bytes is an alpha byte:
smask_size = (png_get_rowbytes(png_ptr(img), png_info(img)) / 2)
* png_get_image_height(png_ptr(img), png_info(img));
Interestingly, texk/web2c/luatexdir/image/writepng.w gets it right:
smask_size = (int) ((png_get_rowbytes(png_p, info_p) / 4) * png_get_image_height(png_p, info_p));
Reproduction instructions:
1. (optional) Install texlive-binaries-dbgsym (for line numbers in
valgrind output).
sudo sh -c 'echo "deb http://debug.mirrors.debian.org/debian-debug/ stretch-debug main" > /etc/apt/sources.list.d/debian-debug.list'
sudo apt-get update
sudo apt-get install texlive-binaries-dbgsym
2. Run valgrind and see errors.
valgrind pdflatex ownCloudClientManual.tex
Note that the valgrind stack trace says write_png_gray_alpha, not
write_png_rgb_alpha, probably as an artifact of optimization.
==13526== Conditional jump or move depends on uninitialised value(s)
==13526== at 0x4C300D3: memcpy@GLIBC_2.2.5 (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==13526== by 0x506E425: ??? (in /lib/x86_64-linux-gnu/libz.so.1.2.8)
==13526== by 0x506EE67: ??? (in /lib/x86_64-linux-gnu/libz.so.1.2.8)
==13526== by 0x506FE53: deflate (in /lib/x86_64-linux-gnu/libz.so.1.2.8)
==13526== by 0x199F99: writezip (writezip.c:71)
==13526== by 0x15215C: pdfflush.part.39 (pdftex0.c:18943)
==13526== by 0x18FA1D: write_png_rgb_alpha (writepng.c:381)
==13526== by 0x18FA1D: write_png (writepng.c:662)
==13526== by 0x18A836: writeimage (writeimg.c:370)
==13526== by 0x16BB78: zpdfwriteimage (pdftex0.c:22285)
==13526== by 0x16D794: zpdfshipout (pdftex0.c:24722)
==13526== by 0x17F65C: maincontrol (pdftex0.c:38501)
==13526== by 0x12F5B9: mainbody (pdftexini.c:5656)
==13526==
==13526== Conditional jump or move depends on uninitialised value(s)
==13526== at 0x4C301EB: memcpy@GLIBC_2.2.5 (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==13526== by 0x506E425: ??? (in /lib/x86_64-linux-gnu/libz.so.1.2.8)
==13526== by 0x506EE67: ??? (in /lib/x86_64-linux-gnu/libz.so.1.2.8)
==13526== by 0x506FE53: deflate (in /lib/x86_64-linux-gnu/libz.so.1.2.8)
==13526== by 0x199F99: writezip (writezip.c:71)
==13526== by 0x15215C: pdfflush.part.39 (pdftex0.c:18943)
==13526== by 0x18FA1D: write_png_rgb_alpha (writepng.c:381)
==13526== by 0x18FA1D: write_png (writepng.c:662)
==13526== by 0x18A836: writeimage (writeimg.c:370)
==13526== by 0x16BB78: zpdfwriteimage (pdftex0.c:22285)
==13526== by 0x16D794: zpdfshipout (pdftex0.c:24722)
==13526== by 0x17F65C: maincontrol (pdftex0.c:38501)
==13526== by 0x12F5B9: mainbody (pdftexini.c:5656)
.. more ..
Reply to: