Re: Improvement on parallel make

[Alexey Neyman]

On Monday 11 December 2006 16:42, Christophe LYON wrote:
> > On Monday 11 December 2006 15:50, Christophe LYON wrote:
> >> I don't think so:
> >> x2 will wait for x1 to complete, and x3 will wait for x2.
> > 
> > I applied it and tried it out:
> > 
> > address@hidden:~/work/make> cat mak
> > .PHONY: a b
> > a: x1 .WAIT x2
> > b: x2 .WAIT x3
> > x1 x2 x3:
> >         @echo start $@; sleep 1; echo finish $@
> > address@hidden:~/work/make> ./make/make -f mak -j a b
> [...]
> > address@hidden:~/work/make> ./make/make -f mak -j b a
> [...]
> 
> I see.
> I knew this "feature"....
> 
> Indeed, when invoked with "b a", b is started first, and as x2 has 
> no wait constraint yet, it starts immediately.
> 
> What would you expect? The same sequence in both cases?

Yes. By the way, BSD make will behave in the same way in both cases. 
It would be confusing to specify that "x1 and x2 cannot run in 
parallel" and have them parallelized.

Regards,
Alexey.

-- 
Your destiny, they claim, lies along
a more subtle though crucial course.
                        -- Supox, SC2