On Tue, 2013-03-12 at 14:23 +0200, Tanu Kaskinen wrote:
> On Tue, 2013-03-12 at 12:13 +0100, David Henningsson wrote:
> > On 03/11/2013 07:18 PM, Tanu Kaskinen wrote:
> > > if (!(m->dl = lt_dlopenext(name))) {
> > > - pa_log("Failed to open module \"%s\": %s", name, lt_dlerror());
> > > + /* We used to print the error that is returned by lt_dlerror(), but it
> > > + * was reported that the error message was sometimes incorrect ("file
> > > + * not found" when the file did exist). No message is better than an
> > > + * incorrect message. It would be nice to find and fix the real bug
> > > + * that caused the incorrect message, but that would require someone
> > > + * who is able to reproduce the bug to investigate the it more deeply.
> >
> > I've seen this happening when we have an arch mismatch, e g pulseaudio
> > binary is amd64 and the module is i386.
> >
> > I've improved the Ubuntu packaging recently so it should no longer be
> > possible to install i386 pulseaudio-module-* packages, if the amd64
> > pulseaudio package is installed. I've asked Sjoerd to do the same for
> > Debian, not sure if he's got to it yet.
> >
> > Reference:
> > https://bugs.launchpad.net/ubuntu/+source/pulseaudio/+bug/1078543
>
> Thanks for the information. I should be able to reproduce this then, and
> make a proper fix.
I found out that this is an old and well-known bug in libltdl. Of
course, there aren't any easy fixes for it, otherwise it would already
have been fixed.
I'll apply this patch after making the comment a bit more informed.
--
Tanu
