Re: GTK+-2.0.0 release schedule

[James Henstridge <james daa com au>]

Miroslaw Dobrzanski-Neumann wrote:

> On Wed, Feb 13, 2002 at 06:35:33PM +0000, Joel Becker wrote:
>
> > On Wed, Feb 13, 2002 at 01:04:51PM -0500, Owen Taylor wrote:
> >
> > > - Collect information about what platforms GTK+-2.0.0 will be expected
> > >   to compile and work on.
> > 	Ok, this is going to illustrate my strange biases, but if anyone
> > has an AIX box with xlc on it, I'll do any work required to verify/fix
> > the 2.0 build on AIX.  I'll even build AIX packages of the binaries like
> > I used to back in the day.
>
> One word about runtime loading and libtool on AIX.
>
> Starting from AIX 4.x there is an dlopen()/.. interface. It works fine. But for
> some obscure reasons runtime loading is only partially supported when not
> using the -brtl link option wich is available first on AIX 4.3 (-brtl does not
> work with DCE); libtool does not use the -brtl option for linking. The
> partially support especially means that dlsym() for a symbol located in main
> module (executable) will not work.
There seems to be something in libtool.m4 that mentions -brtl.  It 
checks to see if -brtl occurs in LDFLAGS, and if so sets 
aix_use_runtimelinking to yes (which it then checks while setting 
various link commands).

If libtool is not working for AIX, then you should report it as a bug 
(assuming you have tested with the latest libtool release).

> So the gmodule test fails because it try to lookup a symbol in executable.
>
> * If any component stick to g_module_open(NULL, ...) it will not work on AIX.
>  h = g_module_open(NULL, ...);
>  sym = g_module_symbol(h, "anything", ...);
>  sym is always NULL
>
> * Could module-test.c be rewritten not to use own symbols?
I think the aim of these tests is to see if those APIs work.  Removing 
them sounds like it would defeat the purpose of having a test suite 
(especially if the -brtl flag you mentioned would make the test pass)

James.

--
Email: james daa com au
WWW:   http://www.daa.com.au/~james/