392334 - Using system cairo when --with-system-cairo not specified causes Segfault (SEGV) on viewing SVG file

Status: RESOLVED INCOMPLETE

(SeaMonkey :: Build Config, defect)

Description

[max baker]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.6) Gecko/20070814 SeaMonkey/1.1.4
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.6) Gecko/20070814 SeaMonkey/1.1.4

When making my own build to run out of $HOME to get around not having root I experience segfaults when accessing the cairo library such as viewing an SVG.

I tracked this down to the fact that seamonkey is building it's own cairo but is using the share library in $HOME/lib during runtime, and they are different versions and builds. 

I fixed this simply by adding --with-system-cairo.    

I think the intended behavior should be to imply --WITHOUT-system-cairo as the default and make sure that if there is a libcairo in the ldpath that it not be loaded.

Reproducible: Always

Steps to Reproduce:
1. Have a system cairo in your LD_LIBRARY_PATH
2. Compile version of seamonkey that does not specify --with-system-cairo
3. Run seamonkey and load up a .svg file.  
Actual Results:  
Segfault

Expected Results:  
No Segfault.  Should use libcairo inside of seamonkey build, not system.

you guys rock... keep up the good work.

Comment 1

[Tony Mechelynck [:tonymec]]

I don't compile SeaMonkey so I can't test this.

Max, do you still see this:
- with Sm 1.1.9 ? (current release build)
- with Sm 2.0a1pre ? (current trunk build)

Comment 2

[Peter Weilbacher]

Max, what evidence do you have that system cairo is used even when the in-tree version is compiled and linked in? If I remember correctly cairo gets statically linked into the layout library (components/libgklayout.so on Linux) in SM 1.1.x, I don't see how it could still use system cairo symbols.

Comment 3

[timeless]

please *attach* your complete mozconfig (or configure equivalent).

also output from ldd seamonkey-bin {components,}lib{xul,gklayout}.so as an *attachment* would be appreciated.

Comment 4

[max baker]

Hmm, this bug was from 8 months ago.  I'm afraid I no longer have that system to work with and will not be able to regression test this for you.

Comment 5

[Peter Weilbacher]

Thanks for the feedback, Max. So I guess this will remain incomplete.