Re: buildworld FAILURE:googletest

From: Simon J. Gerraty <sjg_at_juniper.net>
Date: Wed, 30 Oct 2024 22:59:51 UTC
Lexi Winter <lexi@le-fay.org> wrote:
> > meta mode makes very little difference for a clean tree build,
> > it is only update builds where the .meta files become super useful.
> 
> this was a build updating an existing obj tree from
> 
> ~53314e34d5e8e7f781ab990805b22f7a56bc0580
> 
> to
> 
> ~22429a464a5f4f6bb5a056aae1353985db83b721
> 
> so about 10 days worth of changes.

Ok so META_MODE should have been useful.
 
> > Though they are always good for build failure triage - helps if you have
> > MK_META_ERROR_TARGET=yes then you would get the failing .meta file(s)
> > copied to ERROR_LOGDIR which defaults to ${SRCTOP:H}/error/
> 
> i will look at setting this in case it helps debug any future build issues.

If nothing else it eliminates the ambiguity.
If you have a error/meta*.log you *know* that is where your build
failed.

I ran a CI builder for about a decade, which about 90% of the time could
id  the commit that broke the build.  The above simplified its logic
considerably.

Of course you need to rotate error each time (should probably add that
to FreeBSD - contrib/bmake/mk/newlog.sh is used)