-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Danny van Dyk wrote:
> Am Montag, 11. September 2006 03:44 schrieb Zac Medico:
>> Hi everyone,
>>
>> Portage currently has two hard-coded lists of variables that control
>> the behavior of env-update.
> The same applies to eselect env.
> 
>> I'd like to make these variables 
>> configurable so that package maintainers have direct control over
>> them.  The variables break down into two basic types: colon
>> separated and space separated.
> Wrong, there is another type of var, namely those which are not 
> cummultative(sp?).

Okay, then how about if COLON_SEPARATED and SPACE_SEPARATED are
represent cumulative variables, while all other variables are
considered non-cumulative?

>> What is the best way to propagate information about these two
>> variable types?  For example, we can have a list of variable names
>> stored in a new variable called "COLON_SEPARATED" that will reside
>> in either the profiles or /etc/env.d/ itself.  Variable names not
>> listed in COLON_SEPARATED can be assumed to be space separated.
>> Does anyone have any ideas to share about how this information
>> should be propagated?
> a) You'd need two variables to store that info, as we have 3 classes
>    of envvars.
> b) This is in no way related to the package tree. Those vars don't
>    magically change their class from colon separated to non
>    cummultative. I'd like to keep it hardcoded, as loading of such
>    variables can go wrong and you'd end up without a working env tool.

We can store them in /etc/env.d/ itself.  The env-update tool could
be hare coded to consider COLON_SEPARATED and SPACE_SEPARATED as
being implicitly within the SPACE_SEPARATED class.  The tool would
make one pass to accumulate those two variables, and then another
pass to process the rest of the variables.

Zac

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)

iD8DBQFFBYVD/ejvha5XGaMRAojeAKDxFNYaAk668XhmXVjELQVl5PppsACfX0/n
Wmlxhwj+taaQ0H/aXoIswXI=
=mLCw
-----END PGP SIGNATURE-----
-- 
gentoo-dev@gentoo.org mailing list

Reply via email to