Subject: Re: [change request] pattern for patch filenames
To: Roland Illig <email@example.com>
From: Chris Jepeway <firstname.lastname@example.org>
Date: 07/09/2004 15:17:52
0.02USD from a pkgsrc user:
>> I would like the pattern for patch filenames to be changed. In
>> general it is the shell pattern "patch-*", but there are two
>> occurrences of the AWK pattern "patch-[A-Za-z0-9]+". I would like to
>> have the underscore ("_") included in the allowed letters.
Isn't the problem here that awk and shell aren't matching the same
And not really about the _? If so, why not just "patch-.*"? Sure, the
awk pattern matches more strings, but it won't match more filenames if
those names don't contain directory components. There's always more
regexp power a la "/patch-[^/]*$" if it's needed.
And, FWIW, in the month-old version of pkgsrc I use,
there are some Perl patterns that'll need fixing.
> Naming scheme for patch files:
> - grant would rather name the patches by function
I prefer the functional naming approach, too.
It's easy to delete a patch that doesn't happen
to apply if you don't need that function. Otherwise,
you've got to edit every patch file that's adding
strlcpy() support, eg.
> The other point (the naming scheme) is important, too. I noticed that
> with encoding the file name as the patch name the order in which the
> patches are applied only depends on the file name, not on the package
> maintainer's opinion.
Hm. Not "opinion" but "discretion." If a source file can be
patched by more than one patch file, it's useful to be able to apply
them in a specified order. You're in this situation when, eg,
you're merging in patches from other *BSDs, and you don't want
to cat the two patches into one.
> There might be problems with dependent files, such as ./configure and
> ./configure.ac, as well as lex and yacc files.
Glob order (and the discretion of the patch maintainer) to the rescue...