On Monday, April 21, 2003, at 09:02AM, Mike Barcroft <mike_at_FreeBSD.org> wrote: >David Leimbach <leimy2k_at_mac.com> writes: >> I guess my point was I didn't know how to make these "conform" when I >> looked >> at the POSIX standard for SUS. All it said about strerror is that any >> int should >> work and return a string per the locale that is currently set. > >I agree with Tim; the main purpose is to prevent regressions in our >implementation, not to evaluate conformance to any given standard. In >fact, I created the test to ensure I didn't break strerror() or >strerror_r() while fixing a conformance bug in strerror_r(). Well yes... that is what a regression would be :). But I think a standards compliance test suite wouldn't hurt either :). Whether or not we regress doesn't matter much if we don't conform to the standard to begin with... it just means we haven't changed behavior. > >> > cc -O -pipe -mcpu=pentiumpro test-wmemset.c -o test-wmemset >> > for p in test-strerror test-wcschr test-wcscmp test-wcslen >> > test-wmemchr test-wmemset; do >> > /home/tim/p4/wchar/src/tools/regression/lib/libc/string/$p; done >> > PASS strerror() >> >> Failed for me due to failing assertions. > >What's the output of: > >ident /usr/lib/libc.so.* | grep strerror.c > I am not at that machine at the moment. I just did a new buildworld/installworld before leaving for work today. I can check this at lunch. Dave >Best regards, >Mike Barcroft > >Received on Mon Apr 21 2003 - 05:35:59 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:04 UTC