Re: Issue 1201 in lilypond: GUB regtest building flakey?

[lilypond]

Updates:
        Summary: GUB regtest building flakey?
        Owner: joeneeman
        Cc: n.puttock
        Labels: -Type-Build Type-Defect

Comment #1 on issue 1201 by percival.music.ca: GUB regtest building flakey?
http://code.google.com/p/lilypond/issues/detail?id=1201

Running lilypond from
  target/linux-x86/root/usr/bin/lilypond
produces a good bookparts.ly

Is it possible that there's a memory leak, or some unitialized  
variable/pointer, in the spacing code?  In particular, something that was  
added after 2.13.28 ?  That could explain why GUB seems to screw up the  
page counts for regtests when running in a huge collection, whereas  
individual runs work.

Alternately, could there be any information leakage between multiple  
processes?  AFAIK, GUB calls lilypond with -j8 CPU_COUNT=8.


... aha!  I've reproduced it with linux-x86:
  rm -rf input/regression/
  ../lilypond/configure
  cd input
  nice make -j8 CPU_COUNT=8 doc

I end up with 52/lily-23a75a14.png  showing wonky page numbers (ToC: 1 7 3  
4), whereas running lilypond lily-23a75a14.ly produces good stuff (ToC: 1 2  
3).


I'm moving this to a Defect since I've reproduced it without GUB.  note: I  
have *not* applied Joe's lilypond-book fix.  I don't think that has  
anything to do with this problem (since it didn't exist in .28), figured I  
should state that anyway.