Re: `make check' overworks one core on my Core2 quad

[David Kastrup]

Graham Percival <address@hidden> writes:

> That could be what's happening... not necessarily parallel
> lilypond-book runs in the same directory, but building the english
> docs + translations at the same time with lilypond-book in
> parallel?
>
> Is there any way to turn "make -jX doc" into
> "make -j1 CPU_COUNT=X doc" automatically?  I've already seen that
> "make -j1 CPU_COUNT=X doc" works, and seems to take the same time
> + have the same CPU load as doing -j4 CPU_COUNT=4.
>
>
> No, I haven't looked at how make processes the -j options, and I'd
> rather not have to do this if I don't have to.  If somebody can
> look into this, I'll continue to look into why GUB can't build the
> 2.12.3 docs.  Deal?

I don't see that one can change the -j invocation after the fact: it
will at the very least hold for the top Makefile invocation.  From the
man pages, it appears doable when using GNU make to pass down -j1
instances as well as CPU_COUNT to _subordinate_ Makefile processes.

It may be feasible to make an artifical target jobcount that will only
be active when -jx has been used and which will recursively call make
with -j1 CPU_COUNT=x.  Since that will be the only action, that -jx is
active in the top invocation would do no harm as long as one does not
call several make targets at once.

All this would not port to other variants of Make, and it might be more
prudent to just bomb out on -jx and explain that CPU_COUNT should be
used instead.

-- 
David Kastrup