Re: wip/mk/git-package.mk cleanups

Aleksej Saushev <[email protected]> Sun, 04 May 2014 11:51:54 +0400
Newsgroups gmane.os.netbsd.devel.packages,gmane.os.netbsd.devel.pkgsrc.wip.general
Message-ID <[email protected]>
Richard Hansen <[email protected]> writes:

> On 2014-05-03 11:50, Aleksej Saushev wrote:
>> +#		If ${WRKSRC} doesn't exist by the time post-extract is
>> +#		run, ${WRKSRC} will be created as a symlink pointing
>> +#		to ${GIT_MODULE.${GIT_REPOSITORIES:[1]}}.
>> +#
>> 
>> I'm afraid, it doesn't clean anything up,
>
> Sure it does:  It eliminates the need to set
> WRKSRC=${WRKDIR}/${GIT_MODULE.foo} in the package Makefile.

Symlinks don't clean anything up, they complicate things rather,
even though it isn't clear from experience on platforms like NetBSD.
In particular, some FreeBSD tools (e.g. pax that is actively promoted by
POSIX and pkgsrc) behave subtly differently in presense of symlinks.
Please, don't introduce them here. At least not until consequences are
well understood.

>> it only introduces more divergence to other <vcs>-package.mk.
>
> OK, I will make similar changes to the other *-package.mk makefiles.
>
>> 
>> -# Cache support:
>> -#   cache file name
>> +# cache file name
>> 
>> This is outright wrong. First, it removes the only way not to cache
>> local repositories,
>
> Any user or package that overrides an internal undocumented variable in
> order to disable caching is asking for trouble.
>
> If we want users to be able to disable caching, then there should be a
> clear, documented mechanism.  I'll work on adding such a mechanism and
> send out a new version of the patches.

There is clear mechanism, even though it isn't documented.
It works well enough to me.


-- 
HE CE3OH...