Created attachment 22695 [details] said pdf Valgrind says ==30025== Conditional jump or move depends on uninitialised value(s) ==30025== at 0x5F6122C: cmsLinearInterpFixed (in /usr/lib/liblcms.so.1.0.16) ==30025== by 0x5F6BC1F: cmsEvalMatShaper (in /usr/lib/liblcms.so.1.0.16) ==30025== by 0x5F71BA6: (within /usr/lib/liblcms.so.1.0.16) ==30025== by 0x4F12315: GfxICCBasedColorSpace::getRGB(GfxColor*, GfxRGB*) (GfxState.h:167) ==30025== by 0x4F12395: GfxICCBasedColorSpace::getGray(GfxColor*, int*) (GfxState.cc:1546) ==30025== by 0x4F6FA01: SplashOutputDev::updateFillColor(GfxState*) (GfxState.h:1162) ==30025== by 0x4EF74BF: Gfx::doTilingPatternFill(GfxTilingPattern*, int, int) (Gfx.cc:1837) ==30025== by 0x4EF7F1C: Gfx::opStroke(Object*, int) (Gfx.cc:1573) ==30025== by 0x4EF06BC: Gfx::go(int) (Gfx.cc:643) ==30025== by 0x4EF4BE5: Gfx::display(Object*, int) (Gfx.cc:612) ==30025== by 0x4F384DF: Page::displaySlice(OutputDev*, double, double, int, int, int, int, int, int, int, int, Catalog*, int (*)(void*), void*, int (*)(Annot*, void*), void*) (Page.cc:442) ==30025== by 0x402327: _ZL13savePageSliceP6PDFDocP15SplashOutputDeviiiiiddPc (pdftoppm.cc:134) When running over the attached pdf
Koji can you have a look?
I coudn't reproduce this. I did followings on Fedora 7. valgrind utils/pdftoppm bug117919.pdf But, it didn't produce "Conditional jump or move ...". Please tell me what environment you did it.
I'm using Ubuntu 8.10 valgrind 3.3.1 lcms 1.16 poppler from git trunk And yeah, doing valgrind utils/pdftoppm bug117919.pdf gets me the conditional jump or move warnings Just wondering, how did you know it was bug117919.pdf?
> I'm using Ubuntu 8.10 > valgrind 3.3.1 > lcms 1.16 > poppler from git trunk > > And yeah, doing > valgrind utils/pdftoppm bug117919.pdf > gets me the conditional jump or move warnings > I did it in the same environment as yours, but it didn't produce the conditional jump or move warnings too. Do you have a display profile file (display.icc) in ~/.xpdf/ColorProfiles or /usr/share/ColorProfiles ? > Just wondering, how did you know it was bug117919.pdf? It's the pdf file attached here.
No i don't have any color profile. About the name strange, my browser suggests me the name attachment.foo not bug117919.pdf
Sorry, I did it with a wrong LD_LIBRARY_PATH. I got the conditional jump or move warnings. I'm investigating this problem. > > About the name strange, my browser suggests me the name attachment.foo not > bug117919.pdf > Use external viewer or save it instead of viewing it with a plugin and you will see the file name "bug117919.pdf".
Created attachment 22846 [details] [review] the patch fix a bug about cms profile cache First, I've found that the cache shifting writes out of the cache array. This sometimes makes poppler crash. I attached a patch fix this bug.
Created attachment 22847 [details] [review] patch fixes this problem When stroking with a pattern, set strokeColor with a copy of fillColor. This is wrong and the fillColor may be uninitialized. This is the cause of "Conditional jump or move depends on uninitialised value". I attached the patch fixes this.
Created attachment 22936 [details] [review] a patch fixes another bug I've found a bug around this codes. The page 1 of bug117919.pdf has a dashed grid. but poppler displayes non dashed grid. This is because that stroke-width is not reset to 0 before pattern is drawn. I attached the patch fixes this bug. # you know that CairoOutputDev doesn't support stroke-with-pattern # because of lack of clipToStrokePath.
I've applied patch from comment #7, i'll wait for patches in #8 and #9 until i am able to merge patches in bug 19670 and bug 19994 as they also touch similar parts of Gfx.cc and maybe they also fix it.
I've commited both patches now, great work :-)
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.