Error report.

Re: Error report.

On 02/13/2012 01:22 PM, John Smith wrote:
> =========================================
>    GNU M4 1.4.16: tests/test-suite.log
> =========================================
> 
> 1 of 116 tests failed.  (14 tests were not run).

Thanks for the report.

> FAIL: test-readlink (exit: 134)
> ===============================
> 
> test-readlink.h:41: assertion failed

This is due to a known POSIX-compliance bug in the kernel/glibc
combination you are using, and has since been worked around in upstream
gnulib.  The next m4 release will not suffer from this test failure.

Meanwhile, you can use the just-built m4 without issues, as it does not
tickle the particular readlink compliance bug.

--

-- 
Eric Blake   eblake <at> redhat.com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

Fwd: AIX failed two tests

Re: Fwd: AIX failed two tests

[adding bug-gnulib]

On 02/26/2012 11:54 AM, Michael Felt wrote:
> AIX 5.3 TL 07, vacpp v7 compiler
> 
> CFLAGS=
> CC=
> 
> ./configure --prefix=/opt --mandir=/usr/share/man
> 

> 
> FAIL: test-link (exit: 134)
> ===========================
> 
> test-link.h:139: assertion failed
> 

> 
> FAIL: test-vasprintf-posix (exit: 134)
> ======================================
> 
> test-vasprintf-posix.c:3541: assertion failed

Thanks for the report.  These failures are both in gnulib code, and
gnulib has had quite a few improvements since m4 1.4.16 was released;
perhaps one of these improvements has worked around an AIX issue.
Meanwhile, upstream gnulib documents that AIX 6.1 and newer are
occasionally tested, while AIX 5.1 and newer (including your 5.3) is
rarely tested; I wouldn't be surprised if the problems still persist in
upstream gnulib because your platform is so old.

At any rate, if you are willing to help contribute to gnulib to port to
your platform, we can follow up on the gnulib list on how best to go
about addressing these test failures.  And in the meantime; it's
probably just fine for you to install your just-built m4, in spite of
the test failures, as m4 probably won't tickle those particular bugs.

--

-- 
Eric Blake   eblake <at> redhat.com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

bug in $(wildcard) with trailing slash

Make fails to restrict output to just directories when a wildcard
contains both a trailing slash and internal slashes, even though it does
the right thing with no internal slashes.

$ mkdir /tmp/example
$ cd /tmp/example/
$ touch a
$ mkdir b
$ printf 'all:\n\t <at> echo $(wildcard */)' | make -f -
b/
$ printf 'all:\n\t <at> echo $(wildcard ./*/)' | make -f -
./b/ ./a

Both invocations should have omitted a, since the trailing slash cannot
match a non-directory (other than a symlink to a directory).

$ make --version | head -n2
GNU Make 3.82
Built for x86_64-redhat-linux-gnu

I encountered this while writing a macro for libvirt; as a workaround, I
ended up using a trailing /. to force just directories, then using
$(patsubst) to strip things back to what I wanted in the first place.

cross_dirs=$(patsubst $(srcdir)/src/%.,%,$(wildcard $(srcdir)/src/*/.))

--

-- 
Eric Blake   eblake <at> redhat.com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

Re: bug in $(wildcard) with trailing slash

On 03/02/2012 06:53 AM, Eric Blake wrote:
> Make fails to restrict output to just directories when a wildcard
> contains both a trailing slash and internal slashes, even though it does
> the right thing with no internal slashes.

Sorry about that; I meant bug-make, but my mailer autocompleted to the
wrong list.

--

-- 
Eric Blake   eblake <at> redhat.com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org