Re: Test 52 failure on AIX, HP-UX, Solaris

[Joel E. Denny]

Hi Akim,

On Wed, 3 Feb 2010, Akim Demaille wrote:

> Le 2 févr. 2010 à 10:30, Joel E. Denny a écrit :
> 
> > Here's an attempt at fixing this.  Not yet pushed.  I now need to decide 
> > whether this fix is too much for 2.4.2.  If so, I suppose I could just 
> > comment out the affected test group for 2.4.2, but I'm suspicious that 
> > there's a race condition that could sometimes affect other test groups.  
> > I'd appreciate comments from anyone.
> 
> I don't think it would be much of a problem to put in 2.4.2, after all 
> it's only when something goes wrong that this change is effective.

Good point.  I'll probably put something in 2.4.2, but I need to fix some 
things first.  I discuss that in a separate email.

> I once started a thread on gnulib about getting a module that would 
> replace our subpipe. Bruno pointed out that some of our assumptions in 
> subpipe are wrong.  Maybe 2.5 should aim at using what was done at that 
> time.  ISTR that Paolo made a useful addition to gnulib about this.

Do you mean Bruno's comment here?

  http://lists.gnu.org/archive/html/bug-gnulib/2009-07/msg00045.html

In Eric Blake's reply, he says that Bison is immune to the problem that 
Bruno points out because of Bison's use of m4 diversions.  That makes 
sense to me.  Bruno then argues that this doesn't sound very future-proof.  
That may be true in some applications, but I have a hard time imagining 
Bison's usage of m4 diversions changing any time soon.

So, if I read that thread correctly (I confess that I skimmed), after Eric 
Blake's patches to create_pipe_bidi, the original Tru64 problem that you 
were trying to fix can be handled fine by switching from our subpipe to 
gnulib's create_pipe_bidi.  Given that the Bison bug was reported many 
months ago and that gnulib already offers a likely well tested fix, I'm 
thinking this fix could be part of 2.4.2 as well.  It sounds like you've 
already written the patch.  Could you post it?

I'm thinking of putting off 2.4.2 about another week because of the recent 
bug reports.  I'd really like to be done with the 2.4 series at that point 
if possible.