Re: [GNU Autoconf 2.65] testsuite: 229 failed

[Eric Blake]

[moving to bug-m4; replies can drop bug-autoconf]

According to Gene Spafford on 2/25/2010 11:10 PM:
> Thanks for the reply, Ralf.
> 
> I am enclosing the output of the "make check" on the m4 build so you can see 
> the failures.

M4 failures should be reported to the m4 list; so I've redirected accordingly.

All of the failures in your log look like:

Checking ./006.command_li
@ ../doc/m4.texinfo:979: Origin of test
./006.command_li: stderr mismatch
--- m4-tmp.13684/m4-xerr        Fri Feb 26 01:01:42 2010
+++ m4-tmp.13684/m4-err Fri Feb 26 01:01:42 2010
@@ -1,2 +1,4 @@
+Must be attached to terminal for 'am I' option
+Must be attached to terminal for 'am I' option
 hi
 bye

Each of those tests invoke a subsidiary instance of m4 via the equivalent
to a system() call (that is, an instance of /bin/sh is spawned to find the
subsidiary m4).  My guess is that you have a weird environment setup, such
that /bin/sh attempts to call who(1) during startup and ends up being
quite noisy.  In other words, the bug is not in m4, but in your
/etc/profile, ~/.bashrc, or some other sort of script startup file.  It is
bad practice to have noisy startup environment files.

I know of no way to make m4 robust against such a bad environment, where
the mere act of starting /bin/sh outputs junk to stderr before the
intended command passed to system() is even started.

-- 
Don't work too hard, make some time for fun as well!

Eric Blake             address@hidden

signature.asc
Description: OpenPGP digital signature