Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: ghostscript and jemalloc

> > Potentially is is a bug with ghostscript and gets exposed by
> > jemalloc. Simon found a similar bug in pax recently. It seems that
> > jemalloc causes more activity on freed memory.
> Yes, I'm quite willing to believe that the bug is in ghostscript and I 
> have an open bug about it at (#689711) but it 
> would be useful to confirm that other people are seeing the bug.

You may wish to link gs with efence (pkgsrc/devel/electricfence) and
run the resultant binary; when I try your sample with valgrind on
linux, I get a buncha worrisome results:

==28711== Conditional jump or move depends on uninitialised value(s)
==28711==    at 0x41A55D8: (within /usr/lib/
==28711==    by 0x41A6B71: (within /usr/lib/
==28711==    by 0x41A5FA6: deflate (in /usr/lib/
==28711==    by 0x41A4E91: (within /usr/lib/
==28711==    by 0x413C9BA: s_process_write_buf (in /usr/lib/
==28711==    by 0x413D911: s_filter_close (in /usr/lib/
==28711==    by 0x413CF95: sclose (in /usr/lib/
==28711==    by 0x42BC4F7: (within /usr/lib/
==28711==    by 0x42B951C: pdf_open_contents (in /usr/lib/
==28711==    by 0x42BAADC: pdf_close_contents (in /usr/lib/
==28711==    by 0x42978AC: (within /usr/lib/
==28711==    by 0x4299630: (within /usr/lib/

Which leads me to think the problem lies squarely with gs, not
phkmalloc. :-|

Home | Main Index | Thread Index | Old Index