Charles Bailey <[EMAIL PROTECTED]> writes:
> Charles Lane <[EMAIL PROTECTED]> wrote
..discussing printing of VMS error messages when Perl calls exit()...
>
> | I'd think the default should be "unixoid"...much like the situation
> | with vmsish/unixish times, but would like to hear if anyone is really
>
> My uneasiness arises in part because I don't like to depart from normal
> vmsish behavior (no message on success, message on failure; we'll convenently
> ignore BACKUP :-)) in parts of Perl that interact with the rest of VMS.
> It'd be nice for Perl to provide an alternative to callers having to
> Set Message/NoSev/NoFac/NoId/NoText when they want a silent exit, but
> it seems better that it behave by default like most other VMS commands,
> especially when the caller can't change it short of modifying the
> Perl code itself (well, they could probably say
> C<perl "-Mvmsish=messages" prog>).
True, that's why I was trying to get a feel for how much error messages
are really *used*. It may well be that the bulk of users say "turn them
off, they're just noise!". Or "keep them, I use them!".
Note that turning off the messages with
no vmsish messages;
in all the Perl scripts when running the test suite is rather a
pain....but doing the
$ Set Message/NoSev/NoFac/NoId/NoText
*does* seem to be inherited by subprocesses, so a change to
vms/test.com would help the problems with VMSish error messages in
the test suite.
(should have thought of that earlier...but we get an extra feature out of
it, so that's good)
...philosophical aside on the importance of the test suite...
While it's generally true that the test suite shouldn't dictate the
functionality, we're in a somewhat peculiar situation...it's often
hard to tell if our functionality is correct, and lacking a Unix
system I often look to the test suite to help tell me what "correct"
behavior really is.
Many of the test errors encountered are bogus, simply caused by tests
that don't know how to handle VMS peculiarities, but some are not. To
the extent that we have to have "VMS-specific" modifications to the
tests, they're going to break when tests get changed, and get ignored
along with real problems. I dread getting new versions of Perl going,
simply because it means a lot of work dredging through the tests to
find which errors are real problems and which are not.
So I think it's a worthwhile goal to get the VMS-specific bits *out*
of the test suite, by making the Perl behave in as standard a fashion
as practicable and useful on VMS. We can now argue the exact meaning of
"practicable and useful", if anyone wants :-)
And hopefully, when we see test errors, they'll be few and mean that
something really needs to be fixed, not that J. Linux Hacker changed
a test without suitably VMSifying it.
--
Chuck Lane [EMAIL PROTECTED]
