>>>>> On Sat, 20 Apr 2002 13:17:07 -0400, Michael G Schwern <[EMAIL PROTECTED]> said:
> On Sat, Apr 20, 2002 at 04:03:54PM +1000, Ken Williams wrote:
>> >but skipcheck() really returns two array refs, one to a list of files
>> >found and one to files missing. It doesn't list the files skipped due
>> >to your MANIFEST.SKIP.
>>
>> Even that documentation is ambiguous. If you have a directory
>> named "foo/" with 8 zillion files in it, and you have "^foo" in
>> your MANIFEST.SKIP, is it supposed to report all 8 zillion files
>> in the list of skipped, or can it merely report the directory?
>> I vote for just the directory.
> Ok, but didn't we just finish removing the troublesome code that did
> that?
The implementation was flawed, not necessarily the idea behind it.
Ken's problem is a real one, albeit degenerate. The reason why this
problem came up, is an underspecification of the semantics of the
MANIFEST.SKIP file. The original specification did only think of
files, not directories.
I understand Ken's plea as a vote to extend the semantics with:
If a regular expression in the MANIFEST.SKIP file matches a
directory, the whole directory is skipped.
Strictly speaking, this breaks backwards compatibility, but I'd guess,
this change (if implemented correctly) would not break any existing
distribution.
No, I'm not volunteering to do it, just trying to mediate.
--
andreas
