Pádraig Brady <[email protected]> writes:
> That sounds a little dangerous, as it may spam large files to the output.
> I've already adjusted the test to output textual info in this case anyway
> as the output didn't need to be NUL terminated for the test.
I don't think it would be an issue for coreutils, but fair point. Many
other projects use that script. We shouldn't potentially surprise them
with huge log files.
>> It would also be nice if tests/Coreutils.pm used 'diff -u' instead of
>> 'diff -c'. I assume most of us are far more comfortable reading unified
>> diffs. That file probably was written when coreutils used CVS, so likely
>> not the case at the time.
>
> Sounds good.
>
> Related to more easily processing logs,
> I've also pushed a tweak to avoid time_fail= in the logs
> so we can more easily search for fail=
Looks good. I always end up searching for the string "fail=1", but I
guess if you want to avoid false positives you can use Perl regular
expressions. E.g., like this:
$ find tests -mindepth 2 -type f -name '*.log' \
| xargs grep --perl-regexp '\bfail=1'
Collin
