[Bug localedata/15831] Bad $expect passed from tst-fmon.sh to tst-fmon

["carlos at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

http://sourceware.org/bugzilla/show_bug.cgi?id=15831

Carlos O'Donell <carlos at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
                 CC|                            |carlos at redhat dot com
         Resolution|---                         |INVALID

--- Comment #3 from Carlos O'Donell <carlos at redhat dot com> ---
(In reply to slicer_ghent from comment #2)
> (In reply to Andreas Schwab from comment #1)
> > Try running the test with /bin/bash instead of /bin/sh.
> 
> That was an inspired suggestion, thank you. It worked, but it wasn't
> necessarily because of running with /bin/bash instead of /bin/sh, but
> instead of /bin/dash.
> 
> First, I changed both the first line of tst-fmon.sh from "#! /bin/sh" to
> "#!/bin/bash" and the symlink at /bin/sh (originally "-> dash") to point to
> "/bin/bash". Removing the file glibc-build/localedata/tst-fmon.out and
> re-running "make check" resulted in a successful pass of the fmon test
> (glibc-buld/localedata/tst-fmon.out was empty). Other things failed later in
> rt/, but that's a story for another day.
> 
> Then I changed the symlink /bin/sh to point back to dash but left the header
> of tst-fmon.sh as "#!/bin/bash" and re-ran make check. This failed the fmon
> test and resulted in similar (possibly identical, didn't bother to diff)
> output in tst-fmon.out as in previous failures.
> 
> Next, I changed the symlink /bin/sh to point back to /bin/bash and changed
> the header in tst-fmon.sh back to "#! /bin/sh", as it was before I modified
> it, removed tst-fmon.out and ran "make check" again. These conditions gave
> another successful test with failures occurring later in the rt/ tests.
> 
> I suppose this isn't necessarily a bug in glibc then because it appears to
> be a shell issue. Why dash would read a variable in just fine but pass it
> wrong seems odd, but at least you know there is a shell one of your tests
> doesn't play well with.
> 
> Thank you for your help.

There are patches outstanding to glibc to allow it to build with a POSIX shell
and not require bash-specific extensions.

I suggest looking through back months in libc-alpha, finding the patches that
were posted and adding them to the `Pending Reviews' list so we don't loose
track of them and get around to reviewing them:
http://sourceware.org/glibc/wiki/Pending%20Reviews

Note, you need to be in EditorGroup to edit the wiki:
http://sourceware.org/glibc/wiki/EditorGroup

This would go a long way to fixing this problem.

At present we only support /bin/bash, thus closing as RESOLVED/INVALID.

Feel free to open an enhancement request to support just POSIX shell and we can
use that as a holding place to list all of the patches that have been posted
for this kind of support.

-- 
You are receiving this mail because:
You are on the CC list for the bug.