Re: Build failure tooling improvements

[Kevin Bloom <ktnb%posteo.net@localhost>]

Hi,

Jonathan Perkin <jperkin%pkgsrc.org@localhost> wrote:
> * On 2025-03-21 at 11:58 GMT, Michael Bäuerle wrote:
> 
> >Jonathan Perkin wrote:
> >>
> >> It's been encouraging to see some of the proposals recently to improve
> >> quality.  Something I've obviously been very passionate about for a long
> >> time.
> >>
> >> In the past I've done a lot of work, whether that's daily bulk builds
> >> across multiple operating systems, the CI system, scan failure mails,
> >> custom bulk builds, lots of additional checks, etc, but they haven't
> >> always been well received or used as I'd hoped, which has obviously led
> >> to some frustration on both sides.
> >
> >Maybe there was too little positive feedback, but I have always
> >appreciated your work.
> 
> Thank you!
> 
> >To check changes before commit (to pkgsrc), maybe bulk builds should
> >include wip: <https://releng.netbsd.org/bulktracker/build/1966/wip/>
> 
> It's sometimes possible to add specific packages from wip to bulk builds 
> (I've included some in mine for many years, probably too long!), but 
> it's not possible to include all of wip in builds as there's just too 
> many packages in there, and too much breakage (e.g. fundamental logic 
> bugs that cause scan failures).
> 
> >To check options, I try to build every package three times before
> >commit (but only on NetBSD/amd64 with X11):
> >- With default options
> >- With all options disabled
> >- With all options enabled
> 
> Brilliant, thank you!  This kind of testing prior to commit is always 
> incredibly helpful, and respectful of other people's time.

I do the same thing and usually do a quick run through with the
package. My concern, which was made clear a few months ago, was that
even trival updates to some packages cause breakages on other
systems that I don't have the ability to test it on. Even on NetBSD
9.x in some cases. I was greatful that we had bulk builds to point
out these issues. The issue for me is that I don't have any hardware
other than my NetBSD machine to build and test on.

I'm fairly new to the package maintainer lifestyle so beware of my
nativity. Is there any feasible way that we could get build and
running results without having the hardware (virtual or otherwise)
available to us? Obviously, we can get build results with the bulk
builds and I think that contacting the maintainer is a great idea,
too. I'm more concerned about the running of the package.


kev