Re: CVS commit: src/usr.bin/make

[Kurt Schreiner <ks%ub.uni-mainz.de@localhost>]

On Fri, Feb 15, 2008 at 10:12:39PM +0100, David Holland wrote:
> On Fri, Feb 15, 2008 at 08:08:12PM +0000, Christos Zoulas wrote:
>  > Module Name: src
>  > Committed By:        christos
>  > Date:                Fri Feb 15 20:08:12 UTC 2008
>  >
>  > Modified Files:
>  >      src/usr.bin/make: job.c
>  >
>  > Log Message:
>  > Improve dh's fix:
>  >      - make sure i is initialized
>  >      - don't subtract 1 in comparison, add 1 to the other side
> 
> Does that really fix it? If max is 0, then job->curPos and nr must
> also both be 0, in which case gotNL must be false and fbuf cannot be
> set to true, and the comparison you fixed can't be reached. (And the
> value of i isn't used.)
> 
> If the above reasoning is correct, and this change nonetheless alters
> the behavior, then probably job->curPos is coming up invalid and we
> haven't really fixed the problem but only masked it...
> 
> (Also, while I'm kibitzing, the default-case value for i should be
> job->curPos, not job->curPos-1, because it's going to overwrite that
> position with a \0 and it thus shouldn't be a slot that's already
> used. The buffer is sized so there's space.)
Just as a datapoint: make -j5 on my old P4 w/ Hyperthreading enabled
called from build.sh ... distribution still gives
    compile  compat/rec_search.lo
    compile  compat/rec_seq.lo
    compile  compat/rec_utils.lo
      build  compat/libnbcompat.a
install ===> tools/compat
dependall ===> tools/binstall
    compile  binstall/xinstall.lo
    compile  binstall/getid.lo
       link  binstall/xinstall
xinstall.lo: In stall/xinstall
nbmake: Child (11938) status 0 not in table?

with a runaway make that has to be killed manualy...

Kurt