[csw-maintainers] Evaluating Gentoo Prefix

Peter FELECAN pfelecan at opencsw.org
Mon Sep 23 11:56:36 CEST 2013


"Maciej (Matchek) Bliziński" <maciej at opencsw.org> writes:

> 2013/9/23 Peter FELECAN <pfelecan at opencsw.org>:
>>> Why I think it could be feasible? Our current code hardcodes /opt/csw
>>> and other variants of the CSW string in many, many places. To build
>>> into a different prefix, e.g. ${HOME}/myproject of an unprivileged
>>> user, we would have to correct countless places in our build recipes.
>>> The required amount of work could be larger than adopting a different
>>> codebase.
>>
>> This is exactly what I doubt. Because I don't see the relation between
>> changing the prefixes in the recipes and using another code base. I
>> wonder if you are not proposing to use another set of recipes. But you
>> are a nice guy and will clarify, isn't it?
>
> I am proposing an evaluation of another set of build recipes.
>
> The relation between changing the prefix and the code base is that the
> prefix is hardcoded in many, many places in our code base. To find out
> how many places, you can try this:
>
> find ~/opencsw -name Makefile -o -name '*.patch' -o -name '*.diff'
> -exec gegrep -ril '/opt/csw' {} \;

456 occurrences in 139 files

Reviewing these I cannot see how this cannot be solved by replacing
'/opt/csw' by the "prefix" macro. What I'm missing? Aside of tweaking a
little bit mgar.

> If anyone is looking into working on making OpenCSW buildable into
> another prefix, I suggest looking at Gentoo Prefix because they've
> already done it.
>
> If no one is thinking of building OpenCSW into another prefix, then
> this thread can be dismissed.

IMHO being able to build in another prefix is an interesting exercise,
but not vital, and it will eliminate branches in our recipes such as
'NIM'.

However, what's more interesting is to be able to rebuild the complete
set of packages from scratch. BTW, you have given this as a rationale.

Being able to rebuild from scratch is possible with average but dedicated
work. The main obstacle, from my stand point, is the missing
documentation for making a build farm. This brings us to another thread
which is for the moment moribund.
-- 
Peter


More information about the maintainers mailing list