On 07/21/2010 03:14 PM, Eric Blake wrote:
> Seeing as how I haven't uploaded 2.67 yet (any day now, once my upload
> key privileges are in place...), I took the time to run the testsuite on
> cygwin.  I had to kill a hang in one of the parallel tests, since cygwin
> named fifos still misbehave in cygwin 1.7.5, but that's not autoconf's
> problem (we document it in BUGS, at any rate).  But there were also a
> couple of new, unexpected failures.
> 
> 253: AC_LANG_SOURCE example                          FAILED (compile.at:191)
> 254: AC_LANG_PROGRAM example                         FAILED (compile.at:232)
> 
> It turns out that this line (in both tests):
> 
> gcc -E -dD -o - conftest.c || AS_EXIT([77])
> 
> is not portable to cygwin, where gcc currently compiles ./-.exe rather
> than outputting to stdout.  Do we really need the '-o -' in that command
> line?  And should I be reporting the surprising behavior of creating
> -.exe as a gcc/cygwin bug to the appropriate folks?

And even after fixing that, the test still failed for me, because I got
this output:

./compile.at:186: ./configure $configure_options
stdout:
configure: loading site script /usr/local/share/config.site
# 1 "conftest.c"
# 1 "<built-in>"
#define __STDC__ 1
...
./compile.at:191: sed  '1,/# 1 "conftest\.c"/d' stdout
--- -   2010-07-21 16:03:38.344750000 -0600
+++ /home/eblake/autoconf/tests/testsuite.dir/at-groups/253/stdout
2010-07-
21 16:03:38.282250000 -0600
@@ -1,3 +1,134 @@
+# 1 "<built-in>"

Whoops - the fact that I have a config.site installed made for a false
failure (there are two "conftest.c" lines, and the insertion of an extra
line made the sed script stop at the first instead of the intended
second).  Patch coming up soon...

-- 
Eric Blake   ebl...@redhat.com    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to