Re: bin/60016 (make(1) tests failing since sjg MaybeSubMake changes)

["Simon J. Gerraty" <sjg%juniper.net@localhost>]

Martin Husemann <martin%duskware.de@localhost> wrote:
> > This looks like a different issue:
> >
> > +sh: ./.make: not found
> > +*** [indirect-open] Error code 127
> 
> Maybe, but they also report the bogus -J usage.

The error in the test (opt-jobs-internal) which I quoted the above from,
was a result of *not* producing the expected -J error.
That's the purpose of the .make symlink; to trick make into not setting
OP_SUBMAKE.

That test failed because .make was missing.
Which should not be possible if only opt-jobs-internal was running and
only one instance of it was running.

Hence the questions about the context

> 
> > are these tests being run in parallel?
> 
> Not sure what you mean. It is a regular
> 
>         cd /usr/tests && atf-run | atf-report

Is that being launched from a makefile? if so, which one?

The actual tests are run by tests/usr.bin/make/t_make.sh which will
do so sequentially, and thus should not cause that particular
failure.  But perhaps t_make.sh is being run more than once.

> Both unexpected "make" test failures happened on all tests on different
> machines I did run today (with at a glance identical failure output).

Repeatability helps when it comes to triage ;-)