Re: Changing AT_CHECK

[Akim Demaille]

>>>>> "Pavel" == Pavel Roskin <address@hidden> writes:

Pavel> The patch looks good. I have never used "-n". It didn't work
Pavel> fully anyway since stdout and strerr were not checked. I hope
Pavel> you are aware that "testsuite --help" is still suggesting "-n".

Yep.  Now that everybody seems to agree we don't really need it, I can
remove it.

>> dnl Check stderr.  m4_case([$4], ignore, [cat stderr >&5], experr,
>> [AT_CLEANUP_FILE([experr])dnl

Pavel> This is similar (maybe identical) to the old behaviour, but I
Pavel> don't think it's correct. If stderr is to be ignored then it
Pavel> should not be printed.

I agree it seems logical, but my point is that too much information
when something fails is way better than not enough.  Note it uses fd
5, i.e., this is displayed only when --verbose.

Removing this feature is making useless difficult to get the real
output from the tested tools.

Pavel> In most cases stderr should not be ignored, but if it is then
Pavel> it means that something may be printed on stderr and it's
Pavel> Ok. In this case, repeating possible error messages in the
Pavel> output of the testsuite would be rather confusing than helpful.

Would you suggest a different presentation?

(echo "stderr (ignored):" && sed 's/^/| /' stderr) >&5

?