Re: [PATCH] make -jN check fails / unset MAKEFLAGS in pg_regress.c

On 2012-11-30 14:41:14 -0500, Robert Haas wrote:
> On Thu, Nov 29, 2012 at 8:50 AM, Andres Freund <andres(at)2ndquadrant(dot)com> wrote:
> > Hi,
> >
> > Currently "make -jN check" fails during "creating temporary installation"
> > with:
> > make[1]: *** read jobs pipe: Invalid argument. Stop.
> > make[1]: *** Waiting for unfinished jobs....
> > make[2]: warning: jobserver unavailable: using -j1. Add `+' to parent make rule.
> > in install.log.
> >
> > This is due to pg_regress invoking make while being invoked by make
> > itself. gnu make internally sets the MAKEFLAGS environment variable to
> > remember arguments. The problem in this case is that it contains
> > "--jobserver-fds=4,5" which makes the pg_regress invoked make think its
> > running as a make child process.
> >
> > Now the problem obviously can be worked around by using "make -jN &&
> > make check" instead of "make -j16 check" but I several times now have
> > spent time trying to figure out what I broke so it sees sensible to
> > "fix" this.
> >
> > Any arguments against doing so?
> >
> > The attached patch also resets the MAKELEVEL environment variable for
> > good measure. I haven't seen any indication that its needed, but it
> > feelds safer ;)
>
> Seems reasonable to me.
>
> But shouldn't the comment in the patch say "to be certain the child
> DOESN'T notice the make above us"?

Yes. obviously. Stupid, errr, fingers.

Thanks for spotting.

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services