Re: CVS version test failures on cygwin

[Charles S. Wilson]

Tom Tromey wrote:

> > > > > > "Charles" == Charles S Wilson <address@hidden> writes:
>
> Charles> lex3
> Charles> distname
> Charles> pr9
> Charles> pr87
> Charles> subobj5
> Charles> subobj6
> Charles> subdir5
>
> Thanks.
>
> Can you run each of these like this (e.g.)?
>
>     make check VERBOSE=t TESTS='lex3.text'
>
> That might help me find out what is wrong.
> Unfortunately I don't have ready access to a Cygwin box to do this
> myself.

Okay, I've run the verbose checks -- but on the version of automake that 
was built when I had autoconf-2.52 installed (2.52 was also installed as 
I ran the tests).  Recall that in *this* configuration, the failures were:

ccnoco
lex3
libtool2
pr9
pr87
UNEXPECTED PASS: subdir5

So, I ran the verbose tests on THOSE, not the ones in the earlier list. 
I can reinstall autoconf-2.13, rebuild automake, and run the tests 
specified in the earlier list if you want.

Some notes about the results for the current (autoconf-2.52-based) failures:

ccnoco -- reported on other platforms as "works with autoconf-2.13, 
breaks with 2.52"  Same here.

libtool2 -- I don't have libtool installed; this test looks for 
"libtoolize" within the build directory.  Is that significant?  I have 
just finished building libtool (*); I can install it, and rebuilt/retest 
automake if you think that's a good idea.

(*) the libtool build was mostly successful -- it had three test 
failures of its own: dry-run, build-relink, and quote, as well as a 
skipped 'build-relink2' test.  I think this is because not all of the 
windows-dll-isms have been worked out (cygwin is a unix emulation layer 
that runs on MSWindows)

subdir5 -- when I ran the verbose test, I got "Expected Failure" NOT 
"unexpected pass".  I don't know why the results were different between 
the non-verbose and verbose runs.

--Chuck

cygwin.tar.bz2
Description: Binary data