Re: Imported wip/libmicro - Review

[Jean-Yves Migeon <jeanyves.migeon%free.fr@localhost>]

On Tue, 17 Nov 2009 16:46:37 +0300, Aleksej Saushev <asau%inbox.ru@localhost> 
wrote:
> Jean-Yves Migeon <jeanyves.migeon%free.fr@localhost> writes:
> 
>> On Tue, 17 Nov 2009 13:03:21 +0100, Thomas Klausner
>> <thomasklausner%users.sourceforge.net@localhost> wrote:
>>> On Sun, Nov 15, 2009 at 09:59:09PM +0100, Jean-Yves Migeon wrote:
>>>> I just imported libmicro into WIP, and would like you to review it.
>>> 
>>> Doesn't package with USE_DESTDIR=yes on 5.99.22/amd64:
>>
>> Alright, same player, shoot again -
>>
>> Would it be acceptable to add a variable like ${HARDWARE_ARCH} that
>> expands
>> to `uname -m` (or equivalent, depending on the supported OS) in pkgsrc?
>>
>> I am not quite certain that having such things in the Makefile:
>>
>> INSTALLATION_DIRS+= libexec/libmicro/bin-`uname -m`
>>
>> will work as expected, and providing a PLIST for each MACHINE_ARCH seems
>> overkill to me.
>>
>> Comments?
> 
> Why not patch it to install under "libexec/libmicro/bin" instead?

(moving this discussion to tech-pkg, where it belongs)

Under amd64, you could have bin-i386 and bin-amd64 directories.

I can always patch around all occurences of `uname -m` and replace them
with `uname -p`, but this makes me feel as I am patching to circumvent a
limitation.

More generally: you could have an architecture with multiple variants
(sparc => sun4m, sun4u, sun4v, ...), and I'd like to know if there is a
possibility to make such a distinction with pkgsrc.

Cheers,

-- 
Jean-Yves Migeon
jeanyves.migeon%free.fr@localhost