On Wed, Jun 21, 2006 at 10:37:28AM +0100, Neil Williams wrote:
> However, a recent upgrade (I suspect to libtool or gcc) on amd64 has
> meant that I now get architecture-specific lintian errors involving
> rpath <groan>:
>
> $ lintian pilot-qof_0.0.10-1_amd64.deb
> W: pilot-qof: binary-or-shlib-defines-rpath ./usr/bin/pilot-qof /usr/lib64
>
> I want to develop on amd64 primarily (because it's a faster desktop
> machine) - is this a bug in libtool? Can both lintian AND linda
> overrides be architecture-specific?
I don't think this is something to override. rpath being set is a real bug,
and should be fixed. Even if you can't do that (because it must be done in
libtool), that doesn't mean the problem should be hidden by overriding the
warnings.
> Should I include a patch for libtool that tries to remove
> -L/usr/lib/../lib64 or just remove the /../lib64 component? Right now,
> /usr/lib64/ appears to be the same as /usr/lib/ - is this a transition
> that lintian doesn't know about?
>
> I am the upstream developer for both packages - how can I ensure that
> --disable-rpath is actually implemented in the upstream code? I'd rather
> not just fix this in the Debian packages unless it is a Debian-specific
> issue.
I think it is a Debian-specific issue, due to the upstream libtool maintainer
having a disagreement with Debian about rpath. He considers it a feature
which should always be used, we consider it a feature which may be useful in
some special cases (in particular, hand-build libraries installed in
non-standard locations, such as under a user's home directory), but mustn't be
used for normal distribution-installed libraries (which work fine without it).
AFAIK there are several Debian-specific patches for libtool which upstream
isn't going to include. If this is still the case, then I suppose you should
patch Debian libtool, not upstream libtool.
Also, I don't know where you found --disable-rpath. I found it in some
kde-related package's configure script, but it isn't part of standard
configure scripts AFAIK. It would be very useful though. For example, when
I'm cross-compiling, libtool adds an rpath of /usr/${arch}/lib. This is
completely wrong, of course, since that is not where the library will be
installed on the target platform. It is also harmless, since the target will
not have a cross compiler to itself (that wouldn't be a cross compiler), so it
doesn't start using other libraries from it. Still it makes lintian and linda
complain, and rightfully so. Is anyone aware of a good method to work around
this? Or should libtool be patched with cross-compiler paths as well? That
sounds hackish, but it may be the best solution...
Thanks,
Bas
--
I encourage people to send encrypted e-mail (see http://www.gnupg.org).
If you have problems reading my e-mail, use a better reader.
Please send the central message of e-mails as plain text
in the message body, not as HTML and definitely not as MS Word.
Please do not use the MS Word format for attachments either.
For more information, see http://129.125.47.90/e-mail.html
Attachment:
signature.asc
Description: Digital signature