[pacman-dev] makepkg: keep PACMAN_OPTS in environment

[Allan McRae]

On 28/07/11 01:49, Dave Reisner wrote:
> On Wed, Jul 27, 2011 at 10:31:30AM -0500, Matthew Farkas-Dyck wrote:
>> Patch to make makepkg keep PACMAN_OPTS environment variable, thus
>> allowing operation to be better controlled, e.g. by an AUR- or other
>> helper script to build packages on and install packages to a mounted
>> guest filesystem.
>>
>> --- pacman-3.5.3_/scripts/makepkg.sh.in	2011-06-07 10:49:28.000000000 -0500
>> +++ pacman-3.5.3/scripts/makepkg.sh.in	2011-07-27 09:33:11.413361098 -0500
>> @@ -80,8 +80,6 @@
>>   # when dealing with svn/cvs/etc PKGBUILDs.
>>   FORCE_VER=""
>>
>> -PACMAN_OPTS=
>> -
>>   ### SUBROUTINES ###
>>
>>   plain() {
>>
>>
>> Cheers,
>> Matthew Farkas-Dyck
>>
>
> A few things:
>
> 1) We accept properly formatted patches against our git repository:
>
>    http://www.archlinux.org/pacman/submitting-patches.html
>
> The source of makepkg has changed significantly as since v3.5.3, there's
> been 505 commits with 10% of those going to makepkg.
>
> 2) This would be an undocumented feature that should be called out in
> the man page.
>
> 3) Not sure I agree with this change in execution as randomly pulling in
> environment variables can and will cause trouble. I'd rather see this
> implemented as an additional flag to makepkg.
>


Hmm....  I thought we maybe already had this feature...


 > PACMAN="pacman --asdep" makepkg -i
==> WARNING: A package has already been built, installing existing 
package...
==> Installing package python-six with pacman --asdep -U...
sudo: pacman --asdep: command not found


So that does not work...  but it would be probably be a very minimal 
change to make it do so.


But I am still not entirely sure what is trying to be achieved here.  If 
you want to build and install packages on a mounted filesystem, you need 
to chroot to that filesystem.  Otherwise, you a building locally.  So 
using --root as a flag is probably not enough anyway...

Allan