Re: Git checkout/clean and double-buildability.
On Sun, May 05, 2013 at 04:30:05AM +0200, Lucas Nussbaum wrote:
> On 05/05/13 at 03:06 +0200, Adam Borowski wrote:
> > On Sun, May 05, 2013 at 10:00:07AM +0900, Charles Plessy wrote:
> > > Anyway, given that our infrastructure builds binary packages from a fresh
> > > unpacked source package, I would prefer if we keep the compromise that
> > > imperfect "clean" targets are not release-critical problems.
> >
> > Note that for a big majority of packages, double builds work just fine.
>
> [citation needed] :)
>
> > In fact, that's most of what "make" is for: coping with partially
> > complete trees.
>
> I think that it depends on what kind of double-builds you are
> talking about. For binary-only builds, I agree with you. But if you try
> to rebuild the source package after a binary build, I'm quite sure that
> a large number of packages will fail to build.
Yeah, I meant binary-only builds. That's where partial builds are important
-- only C/C++ code will be ccached.
I agree with Charles that for source builds, the "clean" target is often
duplicating work that could be better done by git. Yeah, 3.0 (git) would
solve this nicely.
--
ᛊᚨᚾᛁᛏᚣ᛫ᛁᛊ᛫ᚠᛟᚱ᛫ᚦᛖ᛫ᚹᛖᚨᚲ
Reply to: