[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Build-Depends-Indep: debhelper, utilizing it in debian/rules clean

Roland Stigge <stigge@antcom.de> wrote:
> On Fri, 2003-12-19 at 15:47, Roland Stigge wrote:
>> consider #216747, where a package declares "Build-Depends-Indep:
>> debhelper" and uses e.g. dh_clean in the debian/rules clean target.
>> Build-Depends-Indep is defined to be satisfied in the build,
>> build-indep, binary and binary-indep targets, but not in clean.

>> For the main distribution, "grep Build-Depends-Indep .../Sources|grep
>> debhelper|wc -l" results in 1871 packages. I didn't check all of them
>> but I doubt that many of them also declare "Build-Depends: debhelper" or
>> don't use dh_clean in the clean target. So >>1000 source packages seem
>> to behave like the package in #216747.

> Andreas Metzler wrote:
>> I think you are overestimating the issue. Contrary to policy the
>> buildds do not install Build-Depends-Indep when running
>> debian/rules build
>> so these bugs are found fast.

> What this (and the ongoing whole thread) seems to say to me is that
> buildds don't conform to policy. That's great! But it's not _directly_
> connected to the question I asked (of which severity the bug described
> is).

Buildds require more than policy, #216747 is about a package that does
not even fulfil policy.

> I'm concluding that you want to say to me that as long as the buildds
> have the current "relation" to policy, I shouldn't care any further
> about this issue (but then, #216747 isn't serious).

> Otherwise, please explain.

The main point was that your mail was suggesting that we might have
1000 packages in Debian with a hidden FTBFS bugs and I wanted to reject
that, because bugs like #216747 will be found the first time the
package is autobuilt.[1]

I then went on to explain that even a package with policy-correct
build-depends could FTBS on autobuilders (debian/rules build invoked
without Build-Depends-Indep).

And my personal opinion in this respect is that a FBTFS bug in a
package caused by autobuilders no following policy strictly would be
a serious bug. - Policy simply is wrong, if you required
Build-Depends-Indep for "debian/rules build" the split
Build-Depends/-Indep would be completely pointless, because you _have_
to invoke "debian/rules build" even if you are not building the
binary-all targets.
              cu andreas
[1] Binary-all packages are not autobuilt, but if you build them you
will install both Build-Depends and Build-Depends-Indep anyway,
therefore it is no FTBFS but a cosmetical issue (not following policy
without any real harm).
Hey, da ist ein Ballonautomat auf der Toilette!
Unofficial _Debian-packages_ of latest unstable _tin_

Reply to: