[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: Qt versus threading
From:       Alex Zepeda <jazepeda () pacbell ! net>
Date:       2001-09-06 7:58:56
[Download RAW message or body]

On Wed, Sep 05, 2001 at 08:13:05PM +0200, Lauri Watts wrote:

> I'd like to back this up on FreeBSD (-stable in my case, not -current).

Well I don't see any hope on getting this to work with -stable.  Mainly
there's little if any development going on WRT threads on -stable.

> Getting KDE HEAD to compile with --enable-mt is impossible, with bizarre 
> errors (the compiler claiming that kdeconfig.cpp has no main() for example.)

Hmm.  I didn't use --enable-mt, it automagically picked up and latched
onto qt-mt.

> Getting it to compile without --enable-mt took quite some doing, but did 
> eventually work (thanks to Otto Bruggeman for the help).   We came to the 
> same conclusion as Alex, that adding -lc_r got it at least compiling.

The correct way on -stable is to use -pthread.  I'm pretty sure that the
extra hackery added to -current to depreciate -pthread was not back
ported.

> The result however, was unusable, as klauncher crashes on startup with 
> "mutex_destroy failure" errors (and then crashes again when you try to save 
> the backtraces).  Some of the applications start, but that's all, they crash 
> as soon as you do anything with them.

Have you submitted a pr for this?  I know there were a few things that
triggered asserts within libc_r.  It's still quite rough.  Some output
would likely be more helpful here.

mutex_destroy failure sounds a bit like a KDE end problem, no?  IIRC you
can't recursively lock something w/ FreeBSD, but this poses no problem
with Linux.

> I've discussed this with the KDE on FreeBSD port maintainer, and with some of 
> the developers, nobody is very optimistic this can be made to work, with 
> threading.

Has anyone tried a recent -current box?  Between school, work and fighting
PacBell Internet I haven't been able to do much of anything, let alone
even finish kdelibs (which did die in khtml somewhere).  Either way it's
in better shape than say GnuCash (I'd never seen a configure script
trigger an rtld assert trap before, but I guess there's a first for
everything).

- alex

[prev in list] [next in list] [prev in thread] [next in thread]