On 10/07/11 15:34, Bruno Haible wrote: > You didn't mention a test case, how to reproduce? Therefore I have to ask > you to test a modification. >
Sorry, it'd be something like this. I used the current Emacs trunk version (bzr 106019). bzr branch bzr://bzr.savannah.gnu.org/emacs/trunk git clone git://git.savannah.gnu.org/gnulib.git cd trunk ./autogen.sh ./configure make sync-from-gnulib The resulting 'configure' file contains bogus invocations of shell functions like func_gl_gnulib_m4code_pathmax that do not exist because the corresponding modules (or their ancestors) have been avoided. The test case takes a long time, because of the "bzr branch", so to save you the time I ran it with your patch, and it worked fine. Thanks for improving gnulib-tool. You can see an instance of the bug by looking at http://bzr.savannah.gnu.org/lh/emacs/trunk/revision/106019/m4/gl-comp.m4 (a patch that reflects the gnulib update, so it has a fix) and searching for func_gl_gnulib_m4code_pathmax. Incidentally, I'm seeing a whole lotta output of md5sum in gnulib-tool, which I don't understand the reason for -- for example, somehow dup2 is turning into some md5 gibberish, which you can see at the above URL. However, the code works so I've not looked into it.
