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

Re: Bug#929747: qa.debian.org: Please add cross-buildability in summary



Control: clone -1 -2
Control: retitle -2 qa.debian.org: Please add cross build satisfiability in summary
Control: block -1 by -2

On Thu, May 30, 2019 at 05:22:12PM +0200, Helmut Grohne wrote:
> On Thu, May 30, 2019 at 10:57:52AM +0200, Samuel Thibault wrote:
> > https://tracker.debian.org/ has a link to the cross-buildability status
> > of a package. It'd be useful to have a tick or cross on the
> > https://qa.debian.org/developer.php page, e.g. in the CI/Rep column, as
> > a link to the cross-buildability status, to be able to easily check the
> > status of one's own packages.
> 
> If you do, please consider:
>  * crossqa.d.n only works on unstable. If a package is not in unstable,
>    there shouldn't be a link. (e.g. gcc-9)
>  * crossqa.d.n only fills a page if there is some package built for one
>    of the architectures being tested. Therefore no link should be
>    emitted for indep-only packages. Currently, we test for any non-x86
>    release architecture, so if a package only builds for some x86, it
>    will be 404 as well.
> 
> If some API is missing in the service, please get in touch with me.

I talked with Christoph about the missing APIs and it became clear to us
that publishing cross build status doesn't make sense as long as
satisfiability isn't published. qa.d.o. does compute satisfiability for
a while now:

https://qa.debian.org/dose/debcheck/cross_unstable_main_amd64/

Please integrate a satisfiability status before integrating crossqa.d.n.
50% of packages are cross-unsatisfiable, so this is the big fish.

Once that is done, we should revisit the cross-buildability as we have
failures for roughly one sixth of the archive and patches for 1/14 of
the archive. In other words: Every third cross build failure has a patch
sitting in the BTS already. So for now, just checking whether your
package has a patch is a much better use of your time. Beyond that, more
than half of the patches essentially are "use debhelper".

What we need here is more people working on the difficult issues, not
random maintainers staring at undecipherable cross build failures. What
we also need is maintainers replying to bug reports and converting their
packages to using debhelper. I get the feeling that we're putting
priorities in the wrong order.

I'm sorry if this comes across a little blunt, but some of this looks to
me like wasting people's time.

Helmut


Reply to: