Control: affects -1 src:nautilus On Fri, Dec 19, 2014 at 6:31 PM, Simon McVittie <[email protected]> wrote: [...] > On 19/12/14 15:17, Mathieu Malaterre wrote: >> Just for reference, chromium crashed: >> >> [86573.137553] chromium[2233]: segfault at 0 ip 00007fce4aeeb785 sp >> 00007fff21991b20 error 6 in libjasper.so.1.0.0[7fce4aebc000+4f000] > ... >> dmesg actually shows two types of crashes [for Nautilus]: >> >> [69727.744833] nautilus[11822]: segfault at 7ff85cb2d510 ip >> 00007ffb50d17d87 sp 00007ffb26eaa6a8 error 4 in >> libgdk_pixbuf-2.0.so.0.3100.1[7ffb50d03000+20000] >> [71582.118170] nautilus[13187]: segfault at 0 ip 00007f60f1f56785 >> sp 00007f610c0c55b0 error 6 in libjasper.so.1.0.0[7f60f1f27000+4f000] > > Based on this, and on libjasper.so being implicated in your backtrace, > I think this is at least one bug and one feature request, and possibly > a second bug: > > * bug A: libjasper should not crash when presented with a malformed or > truncated .jp2 file
Just for clarification. The file attached 'white.jp2' is a valid JP2 file. It is enormous but still a valid file. For those reading this, `white.jp2` does contains *lots* of resolution levels therefore generating a thumbnail out of this file should be trivial (requires proper implementation though). > * feature request B: nautilus should run thumbnailers in a subprocess > so that a crashing thumbnailer does not bring down all of nautilus Sorry, I thought that was the case already, when I saw the `clone` function. But reading the backtrace again it seems to be run in another thread. > * bug C: libgdk_pixbuf's JP2 loader might also have a separate crash > bug (or it might be dying from memory corruption caused by bug A, > in which case this is not its fault) I haven't been able to reproduce this one. I /think/ this may be caused by the openjpeg thumbnailer mecanism, but I would need more nautilus knowledge to check that. > I'm somewhat sceptical about the feature request, and it might end up > tagged wontfix with "no, we should fix the decoding libraries instead" > if people agree with my point of view, but I'll give you the benefit > of the doubt and clone the bug. If the source code for nautilus contains code to check for file magic number it may also be possible to discard JP2 completely (post-jessie?), or simply discard JP2 with enormous canvas size. 2cts -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
