On Tue, 5 Sep 2000 [EMAIL PROTECTED] wrote:
> I've seen similar symptoms of extra line feeds in *some* versions of
> VMS or the CRTL in relation to the piping code. While I haven't been
> able to track them to their ultimate source, it looks like the extra
> linefeeds are related to subprocess terminations.
>
> One hypothesis: SYS$ERROR is being handled slightly differently in
> more recent versions... the test routine supresses error messages,
> but somehow a LF is still emitted.
That hypothesis sounds about right to me, although I have not yet come
up with a test for it.
At the great risk of stepping things backwards let me ask (since I never
did get a chance to take a close look at the patch that Craig prepared)
how hard would it be to make the vmspipe stuff either conditional on
compiler version or make it optional with a configure.com question and an
artificial macro like PERL_VMSPIPE? For simplicity I guess we could
assume that we left the descrip_mms.template handling of the vmspipe.com
file in place does this sound possible?
The disadvantage in allowing some vmsperl installation to get away without
a functional vmspipe build I can foresee would include an inability to
write open2() or open3() code that would be guarenteed to work on all
VMSperl installations. On the other hand, since vmspipe is still so new
there are probably a good many installations that do not even have 5.7.0
yet.
So now that the deadline pressure is off, would anyone object to making
vmspipe a little more optional rather than a required part of the build?
Peter Prymmer
