Is it time to start regression testing yet?

[Andre Stechert]

2009/6/9 Ian Eiloart <iane at sussex.ac.uk>:
> --On 8 June 2009 20:29:19 -0400 Jeremy Lavergne <jeremy at lavergne.gotdns.org> wrote:
>
>> There's a difference between a crash and a failed compilation in that the
>> system can't just catch it.  Apple also pays people to look at those
>> crash reports.
>
> Paid or not, if you can't see a problem then you can't fix it. Sure the system can't catch a complilation error, but surely the complier returns an error code if it fails? Certainly MacPorts can determine whether it's worked or not. If it were able to report to a port maintainer that a build failed then the maintainer has a chance of fixing it. With more information (eg, stage of the build, that a particular dependency failed, OS version, compiler version, h/w type, the variant in use), it becomes much easier for the maintainer to determine what's wrong. With reports of successful builds, they can even eliminate possible problems.

+1 "can't see it can't fix it"

Note that we currently allow users to log crash reports using TRAC --
an automated build failure notification just makes it easier for the
user and, if appropriately designed (as Ian said), easier for us too.

Re "can't catch it", ironically, wouldn't "catch" be exactly what's
needed on trunk, line 157, of base/src/port1.0/portbuild.tcl,
analogous to similar invocations in portconfigure.tcl and
portfetch.tcl?

Cheers,
Andre