Craig A. Berry <[EMAIL PROTECTED]> writes:
>Sometime since 5.8.0, glob has stopped working in miniperl on VMS:
>
>
>I'm not sure what's changed lately, i.e., whether it's File::Temp that
>has started requiring Fcntl or something about glob that has started
>using PerlIO_tmpfile, which in turn loads File::Temp. The most pressing
>side effect of this is that various MakeMaker utilities depend on using
>glob in miniperl.

What has changed recently is PerlIO_tmpfile() now calls File::Temp.
Jarkko decided to do that rather than try and pull the logic from File::Temp
into perlio.c
This was in resonce to a report that all the good work in makeing File::Temp
safe etc. would be wasted if everyone switched to PerlIO_tmpfile via

open(my $fh,">+",undef);

>
>It seems to me the more general question is whether it's ok for
>PerlIO_tmpfile to require dynamic loading. Will other platforms never
>need this in miniperl?
>
>If VMS is the only platform that will need PerlIO_tmpfile in miniperl I
>can probably cook up an alternative tmpfile routine for VMS; in fact
>we've already got a Perl_my_tmpfile in vms.c that may just need to have
>its FILE* return value imported to PerlIO*.

As this has only occured very recently then previous PerlIO_tempfile()
(which used mkstemp()) should suffice.

-- 
Nick Ing-Simmons
http://www.ni-s.u-net.com/

Reply via email to