Re: Updating my patches

On Tue, Sep 07, 2004 at 07:33:46PM -0400, Barrie Slaymaker wrote:
> > --- Makefile.PL   (/vendor)   (revision 1048)
> > +++ Makefile.PL   (/local)   (revision 1048)
> > +        'Text::Table'         => 0,
> 
> Text::Table is developer-only, like BFD, so I don't hard require it.
> Perhaps I should have a DEVELOPER=1 parameter to Makefile.PL

Yes, but the problem is perl Makefile.PL; make is not happy without
them installed, when building the help or something else.

> > --- lib/VCP/Filter/changesets.pm   (/vendor)   (revision 1048)
> > +++ lib/VCP/Filter/changesets.pm   (/local)   (revision 1048)
> > -               unpack "w*", $changes->[$change_index]
> > -               if $changes->[$change_index];
> > +               unpack "w*",
> > +               exists $changes->[$change_index] ? 
> > $changes->[$change_index] : '';
> 
> How about this:
> 
>             my @future_indexes =
>                exists $changes->[$change_index]
>                   ?  grep $_ != $r_index,
>                      unpack "w*", $changes->[$change_index]
>                   : ();

Looks logically identical and cleaner.

> > --- lib/VCP/Utils/p4.pm   (/vendor)   (revision 1048)
> > +++ lib/VCP/Utils/p4.pm   (/local)   (revision 1048)
> > @@ -830,7 +830,7 @@ sub p4_set_client_spec {
> >     $self->p4( [ "client", "-i" ], \$client_spec, \my $out ) ;
> >  
> >     die "unexpected stdout from p4:\np4: ", $out
> > -      unless $out =~ /^Client\s.*\ssaved\./ ;
> > +      unless $out =~ /^Client\s.*\s(saved|.*not changed)\./ ;
> 
> When does this occur, --continue?  I'm wondering if we should try to
> avoid changing the client spec in whatever case causes this.  And I'd
> like to build a test case for it, if you can tell me how.

It happen's in SVN::Mirror's test suite, when a single test tries to
mirror from a running p4d twice, unlike the vcp test suites the server
and p4root are cleaned.  The client spec for the 2 sync are generated
from the two pid and thus identical and exist on the server, since vcp
only removes the client spec at END or some very late cleanup IIRC.

This will not only happen in the test suite, but also in svk pperl
mode, as the actual process trying to invoke vcp is the same process
that never ends.

> > -lib/VCP/Filter/logmemsize.pm
> 
> That's now in the repository.  Odd, though, neither logmemsize.pm nor
> swapnfs.c,v were in my local MANIFEST.  What gives?

Looking at the p4web they are indeed there.  Maybe you had some
overlay of p4 client spec?

Cheers,
CLK

pgpFdjNeGJ8c4.pgp
Description: PGP signature