Re: Unbreak Cygwin bash igncr mode with ac_cr

[Ralf Wildenhues]

* Aaron W. LaFramboise wrote on Sun, Mar 22, 2009 at 09:27:03AM CET:
> Ralf Wildenhues wrote:
>
>> Hmm.  There has meanwhile been another proposed patch to fix a similar
>> build failure on DJGPP:
>> <http://thread.gmane.org/gmane.comp.sysutils.autoconf.bugs/6577>
>>
>> Could you try it out, please, without any other patches?
>
> The DJGPP has no apparent effect on my igncr build; it fails to  
> configure, so the testsuite cannot be run.

Hmm, ok.  Can you post the configure failure in that case?

> Whatever is causing the  
> DJGPP problem, it isn't exactly the same thing as the Cygwin bash igncr  
> problem.  I guess they are using some other method to throw away \r's.

>>> Here are the testsuite results as of January 2009's 2.63.252-8aa9.
> ...
>>
>> In (2) and (4), which patch would that be, the one you posted in
>> <http://thread.gmane.org/gmane.comp.sysutils.autoconf.patches/6431>
>> or the one you just posted below (that is, in
>> <http://thread.gmane.org/gmane.comp.sysutils.autoconf.patches/6431/focus=6516>
>> )?
>
> This is with the newer patch I just posted, which I think is closer to  
> what Eric Blake had suggested.

Hmm.  Maybe your second patch is fine.  A shell won't look into the eval
if the code path isn't taken, so there shouldn't be a problem after all.
I'll let Eric comment on this (and the other patch proposed for DJGPP),
he has more knowledge about w32 quirks than I do.

>> For a failure, just post tests/testsuite.log.  For a hang, just run
>> something like
>>   make check TESTSUITEFLAGS='-v -d -x 27'
>
> Thanks for the advice.  Revisiting this issue, it looks like this hang  
> is actually an intermittent hang.  At least, it hung indefinitely twice  
> the last time, and when I run individually, it doesn't hang (but it does  
> fail).  I'm going to try to run the testsuite again, and try a little  
> harder to see if I can get a complete run out of it.

Thanks.  But even hangs are interesting, in that it is interesting to
know where things hang.

> By the way, I've noticed on previous occasions that there are some parts  
> of the autoconf testsuite that randomly fail intermittently on Cygwin  
> every now and then.  I don't know what this is about.  I've noticed  
> Cygwin fork() will sometimes fail for no obvious (to me) reason, so  
> maybe this is it.

And that is interesting, too.  If you have to kill the testsuite, there
will likely be verbose logs in tests/testsuite.dir/NNN/testsuite.log,
with NNN replaced by the test number.  So any additional information you
can give on such issues can only help.

Thanks,
Ralf