On Fri, 1 Sep 2000, Craig A. Berry wrote:
> At 12:03 PM 9/1/2000 -0700, Peter Prymmer wrote:
> > char fname_phdev[NAM$C_MAXRSS+1];
> >......^
> >%CC-E-BADSTMT, Invalid statement.
> >at line number 5605 in file DKB100:[PERL]VMS.C;1
>
> Urk. C 6.2 did not complain, but you are seeing this with C 6.0?
Yes. Odd isn't it? 6.2 had aquired a reputation for being picky.
> >MCR Sys$Disk:[]miniperl.exe "-I[.lib]" [.vms]gen_shrfls.pl -f gen_shrfls.opt
> >%DCL-I-IGNQUAL, qualifiers appearing before this item were ignored
> > \VERSION\
>
> That's because gen_shrfls.pl is invoking a pre-defined cc command and
> appending /VERSION at the end. The compiler tells you it's ignoring the
> other qualifiers when you do that. It's not doing any harm, though it'd be
> nice to fix this in production releases. It's doing the equivalent of this:
>
> $ cc/prefix=all/version
> %DCL-I-IGNQUAL, qualifiers appearing before this item were ignored
> \VERSION\
> Compaq C V6.2-007 on OpenVMS Alpha V7.2-1
Perhaps we should add this:
$cc_cmd =~ s/\/prefix=all//i;
before trying to figure out the gcc version? I'll see if I can test that.
> >The regression tests still skip lines:
>
> Hmm. I'm not seeing this at all. Could the socket stack interacting with
> the piping changes affect this? At the moment I'm testing with Compaq's
> TCPIP services, though I can perhaps try it on a Multinet machine as well.
I was using DECC sockets on a UCX (er, TCP/IP Services), not Multinet
(although I could).
> >BTW Craig, what about stat ing in ODS2 deep directories, ie more that
> >8 levels deep accessed with rooted logicals? Are we sure we want to
> >resolve the physical device name in cando?
>
> The only logical name I'm replacing is the one assigned at mount time and
> that by definition refers to the entire volume. If we encounter a directory
> depth problem, we would have had the same problem using the untinkered-with
> result of lib$fid_to_name.
OK.
Peter Prymmer
