[csw-maintainers] GAR performance, calls to mkpackage

[Dagobert Michelsen]

Hi,

Am 02.10.2011 um 16:56 schrieb Ben Walton:

> Excerpts from Maciej (Matchek) Bliziński's message of Sun Oct 02 09:46:27 -0400 2011:
>> When finishing 'gmake platforms', the build spends quite a bit of time
>> at this step:
>> 
>> * Platform solaris9-i386 (built on host 'unstable9x')
>> Use of uninitialized value $REVISION in sprintf at
>> /home/maciej/src/opencsw/.buildsys/v2/gar/gar//bin/mkpackage line 31.
>> Use of uninitialized value $REVISION in sprintf at
>> /home/maciej/src/opencsw/.buildsys/v2/gar/gar//bin/mkpackage line 31.
>> Use of uninitialized value $REVISION in sprintf at
>> /home/maciej/src/opencsw/.buildsys/v2/gar/gar//bin/mkpackage line 31.
>> Use of uninitialized value $REVISION in sprintf at
>> /home/maciej/src/opencsw/.buildsys/v2/gar/gar//bin/mkpackage line 31.
>> Use of uninitialized value $REVISION in sprintf at
>> /home/maciej/src/opencsw/.buildsys/v2/gar/gar//bin/mkpackage line 31.
>> Use of uninitialized value $REVISION in sprintf at
>> /home/maciej/src/opencsw/.buildsys/v2/gar/gar//bin/mkpackage line 31.
>> 
>> It looks like repeated mkpackage calls.  But gar already knows what
>> packages have been built, doesn't it? Why call mkpackage again?
> 
> It's been a long time since I looked at this part of GAR, but iirc,
> it's using these calls to generate the file names for display.  It's
> relying on some of the gspec magic, I think, and doing it directly in
> GAR was challenging so this made sense.  (This is all from memory.)
> 
> Dago, is this right?

It was mostly due to static gspec files which parsing was a bit difficult and
mkpackage knows best anyway. Now that static gspecs are really deprecated we
could get rid of those. Would you mind opening a bug report?


Best regards

  -- Dago