Bug#459712: build-conflicts with its own runtime

[Kilian Krause]

Faidon,

On Thu, Apr 03, 2008 at 07:29:59PM +0300, Faidon Liambotis wrote:
> [the following is a more useful description of the bug than the irc log]
> 
> pwlib and pwlib-titan build-conflict with their runtime parts besides 
> the -dev parts.
> I think this is being done because a sample program is built and run 
> when the package builds as a mean to catch runtime errors/screwups early.

right. There were some pretty regular cases when subsequent builds were
running fine but executing the lib was then failing with runtime errors.
Therefore this runtime check at build time was neccessary to be introduced.


> I'm not sure, however, if the conflict is really needed.

Unfortunately it has been in the past.


> The current situation is bad because people trying to do build tests on 
> their desktops have to uninstall ekiga among others.

They can still install a fresh chroot (build flavour) which will then allow
them to be clean with the build-requirements that might otherwise demand
them to upgrade the entire system to quite recent versions - and screw up
their daily desktop install. Apart from that it's way easier to control
(security wise) and eventually even wrap up and copy/move around. For these
and other reasons it has not been that much of a problem until now. I have
before and would still even actively discourage any "bug reporter" of this
kind of bugs to do this "native development" and rather go for
chroot/vserver/xen etc.. The ekiga-snapshots have received a
DEB_BUILD_OPTIONS switch to disable this due to some demand. It should not
be a problem to copy this over back to Debian.


> Kilian, you've added both the conflicts and the sample build/run.
> Have you investigated if the conflicts is needed?

Yes.


> In the case that it is, how would you feel about making the sample 
> build/run conditional on the presence of libpt in the build system?

Well, it was very much needed as a precaution in the past - especially on
the non-trivial architectures. Doing a conditional test without enforcing it
to FTBFS the package would just reintroduce that harm with a very low bar to
jump over. If you think it is seriously needed we should add the
DEB_BUILD_OPTIONS switch and conditionalize it in there - or plain not run
it at all.

-- 
Best regards,
Kilian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://lists.alioth.debian.org/pipermail/pkg-voip-maintainers/attachments/20080403/8dfcd91b/attachment.pgp