[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: check4.test consistently failing
From: |
Ralf Wildenhues |
Subject: |
Re: check4.test consistently failing |
Date: |
Tue, 28 Apr 2009 22:42:41 +0200 |
User-agent: |
Mutt/1.5.18 (2008-05-17) |
Hello Philipp,
* Philipp Thomas wrote on Tue, Apr 28, 2009 at 03:22:04PM CEST:
> * Ralf Wildenhues [20090428 08:20]:
> > Hmm. Which version is make, and what shell is /bin/sh?
>
> make is 3.81 and /bin/sh is bash 4.0.10(1)-release.
Hmm. I've tried reproducing your issue with those two programs on a
Debian GNU/Linux, but no luck.
Wait ... have you even shown the full output of the test? Over here,
it will continue after what you've posted, with
+ :
+ cat stdout
Making check in dir
[...]
If that really was the end of the output, then the log definitely looks
like a shell bug, though, as it is failing over the first
$MAKE -k check >stdout || :
line that should never exit, not even after "set -e". Can you try to
debug it yourself? For example, remove the ">stdout" redirection to see
what it outputs, or replace "|| :" with "|| echo $? && echo $?" to see
the return value.
Note that you can rerun this one test only with either
cd tests
make check VERBOSE=yes TESTS=check4.test
or
cd tests
$srcdir/check4.test
where $srcdir is the path to the respective source tree directory
('.' if you are building in-tree). Any differences in the result
between these two methods would be interesting to know, too.
> > Do you have any diffs over FSF Automake 1.10.2 in your tree?
>
> Yepp, both of which I'm attaching. The first one allows to use
> $vendor-linux-gnu instead of unknown-linux-gnu in lib/config.guess and the
> second one fixes require_file_internal to handle file names with
> directories. But AFAICS both don't come into play here.
It is a bit ugly to have config.guess deviate from the canonical
upstream file, which comes from
<http://savannah.gnu.org/projects/config>
Can you please grab the current upstream version of that file, and send
any remaining diffs to address@hidden to discuss it with them?
Second, do you have a testcase that shows what the require_file_internal
patch is required for? Why has this patch not been sent to the
automake-patches list before (or has it, and I've missed it)?
Thanks,
Ralf
> --- automake.in
> +++ automake.in
> @@ -6836,12 +6836,14 @@
> my $fullfile = "$dir/$file";
> my $found_it = 0;
> my $dangling_sym = 0;
> + my $thisdir = dirname ($fullfile);
> + my $thisfile = basename ($fullfile);
>
> if (-l $fullfile && ! -f $fullfile)
> {
> $dangling_sym = 1;
> }
> - elsif (dir_has_case_matching_file ($dir, $file))
> + elsif (dir_has_case_matching_file ($thisdir, $thisfile))
> {
> $found_it = 1;
> maybe_push_required_file ($dir, $file, $fullfile);