[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
gnulib-tool --with-tests --test
From: |
Ralf Wildenhues |
Subject: |
gnulib-tool --with-tests --test |
Date: |
Thu, 9 Nov 2006 00:12:27 +0100 |
User-agent: |
Mutt/1.5.13 (2006-08-11) |
It would be nice if
gnulib-tool --with-tests --test
succeeded out of the box, including actually running the tests. IMHO
it would make a (one) good candidate for automatic testing. Currently
it has (at least) the following issues:
- automake throws a warning about `po' not being mentioned in `SUBDIRS'.
- gltests/ is not autoreconf'ed, thus not configured and built either.
If I manually run
autoreconf -vi
in there, autopoint does not like the fact that its installed files
have been modified (by the bootstrap of the top directory). It seems
AUTOPOINT=true autoreconf -vi
works, except for:
- gltests/configure.ac contains lines of the form
AC_CONFIG_FILES([csharpcomp.sh:build-aux/csharpcomp.sh.in])
that should be
AC_CONFIG_FILES([csharpcomp.sh:../build-aux/csharpcomp.sh.in])
and that lead to automake errors.
- The sub-configure in gltests fails because it looks for getloadavg.c
in the wrong directory:
| configure: error: ../../gltests/gltests/getloadavg.c is missing
| configure: error: ../../gltests/configure failed for gltests
I'd say there is an AC_CONFIG_LIBOBJ_DIR missing, but now that gnulib
does things differently, I don't know how to fix this.
- I would like an option for gnulib-tool so that it does not remove the
resulting directories after the test, and one to omit the 'distclean'.
Manually inspecting the source and build trees helps to uncover more
bugs. For example:
- I get this output from a ../configure -C (in an empty directory):
| checking for working acl_get_file... no
| (cached) checking whether program_invocation_name is declared... yes
| checking whether program_invocation_short_name is declared... yes
| checking whether program_invocation_name is defined... yes
| checking whether program_invocation_short_name is defined... yes
| checking for argz.h... yes
Note the '(cached) ' at the beginning of the line. Not done
investigating yet, but it seems a weird macro invocation vs.
macro AC_REQUIREment issue again...
After things are cached, the output even looks like this:
| checking for working acl_get_file... (cached) no
| (cached) (cached) checking whether program_invocation_name is declared...
(cached) yes
| checking whether program_invocation_short_name is declared... (cached) yes
| checking whether program_invocation_name is defined... yes
| checking whether program_invocation_short_name is defined... yes
[...]
Hope that helps.
Cheers,
Ralf
- gnulib-tool --with-tests --test,
Ralf Wildenhues <=