On Sunday 18 September 2011, Stefano Lattarini wrote:
> On Sunday 18 September 2011, Stefano Lattarini wrote:
> > Severity: serious
> > 
> > The bug is demonstrated by the attached test script.
> > I hope I'll be able to take a look into it soonish.
> > 
> The test script has a couple of bugs, which I've now fixed.  Attached
> is the patch I've pushed to expose the bug.
> 
OK, I've a patch that fixes the bug (see attachment).

But wait to rejoice, since the patch also introduces new failures when
"threaded automake" is in use (AUTOMAKE_JOBS > 1).  The reason is that
automake needs to serialize file installions spawned by `--add-missing'
(and if I understand correctly, this isn't for cosmetic reasons, but is
needed to avoid possible race conditions and botched output files).
Currently the code that installs required files is intertwined with the
code that declares the DIST_COMMON variables; so, upon deserialization,
the definition of DIST_COMMON gets emitted in the wrong Makefile.in
(ouch!).

I hope I'll have time and patience to look into this soonish; anyway,
suggestions on how to minimally fix the new problem would be very
welcome (the best & long-term fix would probably entail a quite heavy
refactoring of automake, which, altough defintely desirable, would
require a lot of time and testing ... so not going to happen soon,
sigh).

Regards,
  Stefano