At 10:08 PM -0500 3/28/02, Michael G Schwern wrote:
>The bit which bothers me is the "adequate for at least 85-99% of
>usage".  It's gotta be 100% (less stupid user input).

OK, I think we're getting carried away about the fact that it would
be nice to have a completely general purpose catfile() equivalent for
the shell but in fact we all we need is a sufficiently robust way to
append directory and filename pieces to the following 4 macros:

PERL_ARCHLIB
SITEARCHEXP
INSTALLARCHLIB
INSTALLSITEARCH

and in some case the piece to be appended includes the string 'auto'
and/or the macro:

FULLEXT

, macros over which we have considerable control.  I think Peter's
suggestion as modified by John to handle both ']' and '>'  is a
reasonable way to go:

<http://www.xray.mpe.mpg.de/mailing-lists/makemaker/2002-03/msg00088.html>

The only ways I can see it causing trouble is if someone configures
an individual macro to use both square and angle brackets (but that's
in invalid filespec anyway), or if someone configures one piece of a
path with square brackets and another piece with angle brackets.  But
guess what, catfile doesn't handle the latter case properly either:

$ perl -"MFile::Spec" -e "print File::Spec->catfile('dev:<dir1>','[.dir2]','foo');"
dev:[<dir1.dir2]foo

So, what are the additional risks of going with the all-DCL method?
I can't think of any assuming we can get the piece that reads in the
temp file fixed to handle newlines.  I will resynch and attempt a new
patch along the lines suggested by Peter and John.


>
>I'll apply Craig's original patch, mod my little changes.  It's slow,
>but it's complete and I know it works.
>http:[EMAIL PROTECTED]/msg05439.html
>
>So I'm going to punt on this one and let you folks decide if Craig's
>is too slow and needs work.  I'll only take a full patch, though.
>
>Sorry, I'm out of time.  Gotta get a feature complete patch to Jarkko
>this weekend.
>
>
>On Thu, Mar 28, 2002 at 10:47:03AM -0500, [EMAIL PROTECTED] wrote:
>> We've presented a non perl solution to you, that should be
>> adequate for at least 85 to 99% of usage (possibnly all),
>> provided that having record boundaries (i.e. line feeds)
>> in .MM_tmp is acceptable to the perl utility that
>> uses it as input, presumably ExtUtils::Install:pm_to_blib(),
>> or somesuch.
>
>If it doesn't, it can be made to do so.
>
>
>> Please note that I cannot run cvs out of the firewall here and would
>> need to test only on ftp-able or http-able "make dist"
>> distributions.
>
>I'll see if I can get tarballs auto-generated and available for you.
>
>
>--
>
>Michael G. Schwern   <[EMAIL PROTECTED]>    http://www.pobox.com/~schwern/
>Perl Quality Assurance     <[EMAIL PROTECTED]>         Kwalitee Is Job One
>Realize this, sweetheart, I'm squeezing my poodle.


-- 
____________________________________________
Craig A. Berry                  
mailto:[EMAIL PROTECTED]

"Literary critics usually know what they're
talking about. Even if they're wrong."
        -- Perl creator Larry Wall

Reply via email to