On 21.06.2007, at 15:02, josé Alberto Suárez López wrote:

> nice idea :)
>
> El jue, 21-06-2007 a las 14:48 +0300, Philipp Riegger escribió:
>> Good day.
>>
>> I though it might be nice to be able to tell gnap_make which tempdir
>> to use. I can think of 2 scenarios where this will make sense:

I'd like to discuss 2 points i'm not quite sure about.

1) keeptemp: Should this be an extra option (-K) or should this be  
triggered by my introduces custom tempdir option (-T)?

2) My option is quite strange since "usually" you give a tempdir  
like /var/tmp and the tool uses a subdir of that e.g. /var/tmp/gnap- 
lsdfnsdf. My approach uses the given dir directly, this is kind of  
not straight forward. Furthermore, if somebody uses a environment  
variable TEMPDIR and -T is not used, the alternative execution path  
is also used.

Some possibilities what to do:

To fix the environment thing, TEMPDIR can be set to '' before parsing  
command line arguments. As an alternative we could use the GNAP*  
namespace and rename it to GNAPTEMP or GNAPTEMPDIR. This would be  
easier than setting all sensitive variables to '' and if somebody  
messes with that namespace, it's not our fault.

To fix 2) we could use TEMPDIR/gnap or TEMPDIR/gnap-VERSIONSTAMP in  
the -T case. anything against this?

Thoughts about 1): If no overlays are used, the tempdir is quite  
small. If overlays are used, then it is bigger, but the data created  
during the snapshot creation is simply the portage snapshot and the  
overlays on top and can be found in the catalyst tempdir. So... the  
big data is never really needed and it is available at another place  
and the small data is the important one and does not hurt much.

Still: Should we introduce a new command like option or create some  
logic when to delete what and when not? I would prefer 2.

Philipp