[Gtkradiant] [Bug 879] Q2 Linux crashes during startup

gtkradiant@zerowing.idsoftware.com gtkradiant@zerowing.idsoftware.com
Mon, 13 Oct 2003 18:33:21 -0500


ttimo@idsoftware.com changed:

           What    |Removed                     |Added
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED

------- Additional Comments From ttimo@idsoftware.com  2003-10-13 18:33 -------
Fixed by stubbing the _fini function. See below for details.

<TTimo> linux wonders
<TTimo> dlclose a .so, get static destructors in your main app called
<MEGASTeP>  eheh
<MEGASTeP> _fini does that I think
<TTimo> yeah there's a _fini in my backtrace, but what the hell ?
<TTimo> would you know how to prevent that absurdity ?
<Primer> don't call dlclose()....?
* Primer runs
<TTimo> Primer: yes, that's the likely solution
<TTimo> it's what it will come down to if no cleaner solution is found
<TTimo> I could provide a dummy _fini in each .so then
<MEGASTeP> fix gcc
<MEGASTeP> or maybe there's a linker option
<TTimo> tried 3.3.1 and 3.3.2 !
<MEGASTeP> man ld
<TTimo> it only happens with binaries built on woody
<TTimo> and running on sid
<TTimo> bins built on sid run fine on sid for instance
<MEGASTeP> bah debian sucks
<TTimo> well if you want, it crashes on RH9 also
<TTimo> on pretty much any glibc 2.3 system I think
<MEGASTeP> there is a ld -fini
* TTimo looks for a --no-fini option
<vogon> you can always set -fini to a function that does nothing.
<TTimo> which is equivalent and more painful build-wise than putting a fake
_fini in my code, yes
<vogon> heh.
<TTimo> going to try that though