[Rpm-maint] [PATCH] rpmbuild: clean up buildroot handling, set a default buildroot

[Jason Corley]

So maybe the better question is what are the ABI and namespace
concerns in terms of actual usage that make the other approach less
desirable?  Since your patch was written last week and JBJ's has
existed in his branch of RPM for six months (and is being used by
larger distros like Mandriva as well as smaller distros like PLD and
cAos) his would be better tested wrt breakage.  What features that are
in use is that patch breaking?  If the answer is none, then what is
the problem with the separate namespace?
Jason

On 2/15/07, Tom 'spot' Callaway <tcallawa at redhat.com> wrote:
> On Thu, 2007-02-15 at 15:30 -0500, Jason Corley wrote:
> > Nobody seems to care about "BuildRoot: /lib" and that is just as
> > efficient at hosing your system as "BuildRoot: /".  If your goal is to
> > fix a greater problem, then I submit that the patches JBJ has had in
> > his RPM for close to six months are much more a) well tested, and b)
> > more effective as a long term solution.  But just in case you forgot
> > the (only?) arguments against that idea:
> >
> >     1) changes the ABI (not that anyone uses librpmbuild, but....)
> >     2) introduces a new macro namespace
> >
> > Either you want to fix the larger problem or you want to solve your
> > problems.  Either it is unacceptable to change past behavior or it
> > isn't.  Either it is ok to make changes that break other's builds or
> > it isn't.  You decide it's your community, but right now what I see is
> > "we're fine making you change, but don't make us change."
>
> People setting idiot buildroots should be publicly beaten with a trout.
> RPM setting an idiot buildroot by default should be fixed.
>
> The proposal here (at least from me) is that we change the behavior of
> RPM such that it sets a default buildroot, which can then be trumped by
> a system macro, which in turn can be trumped by a local spec define.
>
> This changes the ABI. It doesn't affect the macro namespace.
>
> To the best of my knowledge, Red Hat/Fedora isn't setting idiot
> buildroots, nor are we scraping binary bits into rpms. However, even if
> we were doing that, I'd still be in favor of fixing this bug.
>
> So, to summarize:
>
> I think its ok to change past behavior.
> I think its ok to make well thought out, intelligently discussed changes
> that fix bugs.
> If these changes break other people's builds, then we need to identify
> better ways to make these changes, and/or identify workarounds that
> enable their builds to continue functioning.
> I don't see any indication of Red Hat ducking changes in favor of
> forcing others to change.
>
> ~spot
>
>