>--- Forwarded mail from [EMAIL PROTECTED]
>Paul Kinsley writes:
>>
>> I am thinking of using CVS on an IRIX system, but am concerned that
>> the CVS test suite ( make check ) fails. The first failure is on
>> the test importc-8. The failure on importc-8 is related to
>> the dates rolling over.
>Most test failures are caused by broken or obsolete tools rather than by
>actual cvs problems. In the case of importc-8, the problem is almost
>certainly that your ``touch'' command is not Y2K compliant. You can see
>if SGI has a patch that will fix it, or you can use the GNU version of
>touch instead.
I have also had problems not with the regression suite itself but with
the verification of output. It seems that when the test suite invokes
CVS, it merges stdout and stderr and compares the result to a regular
expression. The problem is that the way the bytes are interspersed is
unpredictable, so there is a huge number of false failures.
I understand the need to verify the order in which messages are displayed,
which is the argument made in the comments. But it would seem that such
a requirement cannot be met in practice.
I've experienced this problem on both Apple A/UX (before the turn of the
year) and NetBSD.
I've also discovered a timing condition in which the test suite must run
to completion when both the local time zone and GMT share the same date.
In my time zone, it means I cannot run the suite successfully between
mid-afternoon and midnight.
>--- End of forwarded message from [EMAIL PROTECTED]
