Re: Speeding up l10n/calc-release-status

[Christian PERRIER]

Quoting Cyril Brulebois (k...@debian.org):

> > > partman-zfs is kfreebsd-* only; others are powerpc-specific, which is no
> > > longer a release architecture. Christian, which update would be the best
> > > one to reflect those changes? Add a minus sign after those packages in
> > > packages_list? Or remove them from there entirely?
> > 
> > I'd suggest tomove them to the attic, at leat the PPC ones. Not sure
> > about partman-zfs, though.
> 
> ACK for the attic move for ppc packages, but mu question was more about
> the l10n part of it.

Well, if the packages are moved to the attic and removed from
packages_list, the translation will vanish from the master files at
the next l10n-sync runand therefore will not interfere with statistics.

> 
> I think partman-zfs should stay where it is (out of the attic), but its
> translations shouldn't be taken into account for stats purposes?

Well, we never did that (not counting arch-specific packages in
statistics), which would enforce the idea that some arches are "more
important" than others. But that's of course debatable given that the
number of people installing a kFreeBSD system in anything else than
English is probably near to zero.

For sure, such packages' translations shoudl be marked for levels 3
and higher and shouldn't clutter the stats for levels 1 and 2.

-- 




signature.asc
Description: PGP signature

Re: Speeding up l10n/calc-release-status

[Cyril Brulebois]

Christian PERRIER  (2017-06-26):
> Quoting Cyril Brulebois (k...@debian.org):
> 
> > Current errors are:
> > E: error getting madison info for partman-newworld for 'testing'
> > E: error getting madison info for partman-zfs for 'testing'
> > E: error getting madison info for prep-installer for 'testing'
> > E: error getting madison info for quik-installer for 'testing'
> > E: error getting madison info for yaboot-installer for 'testing'
> > 
> > partman-zfs is kfreebsd-* only; others are powerpc-specific, which is no
> > longer a release architecture. Christian, which update would be the best
> > one to reflect those changes? Add a minus sign after those packages in
> > packages_list? Or remove them from there entirely?
> 
> I'd suggest tomove them to the attic, at leat the PPC ones. Not sure
> about partman-zfs, though.

ACK for the attic move for ppc packages, but mu question was more about
the l10n part of it.

I think partman-zfs should stay where it is (out of the attic), but its
translations shouldn't be taken into account for stats purposes?


KiBi.


signature.asc
Description: Digital signature

Re: Speeding up l10n/calc-release-status

[Christian PERRIER]

Quoting Cyril Brulebois (k...@debian.org):

> Current errors are:
> E: error getting madison info for partman-newworld for 'testing'
> E: error getting madison info for partman-zfs for 'testing'
> E: error getting madison info for prep-installer for 'testing'
> E: error getting madison info for quik-installer for 'testing'
> E: error getting madison info for yaboot-installer for 'testing'
> 
> partman-zfs is kfreebsd-* only; others are powerpc-specific, which is no
> longer a release architecture. Christian, which update would be the best
> one to reflect those changes? Add a minus sign after those packages in
> packages_list? Or remove them from there entirely?

I'd suggest tomove them to the attic, at leat the PPC ones. Not sure
about partman-zfs, though.



signature.asc
Description: PGP signature

Speeding up l10n/calc-release-status

[Cyril Brulebois]

Hi,

It's been an annoyance during all releases: this script processes all
packages in turn, and that takes quite a while. I've tried to make it
run in parallel by defining a function and calling parallel on it, but
it would need a few other functions and variables to be exported as
well, so I've taken an easier route and split the interesting part into
a calc-release-status-helper script, to be called for each and every
package.

Now, calc-release-status will treat packages in a parallel manner, using
as many cores as available, and it now finishes in under 3 minutes on my
d-i development machine, which makes me quite happy. :)

Other consequence: output will no longer be sorted by package name due
to parallelism.


Current errors are:
E: error getting madison info for partman-newworld for 'testing'
E: error getting madison info for partman-zfs for 'testing'
E: error getting madison info for prep-installer for 'testing'
E: error getting madison info for quik-installer for 'testing'
E: error getting madison info for yaboot-installer for 'testing'

partman-zfs is kfreebsd-* only; others are powerpc-specific, which is no
longer a release architecture. Christian, which update would be the best
one to reflect those changes? Add a minus sign after those packages in
packages_list? Or remove them from there entirely?


Note: It's still on my todo list to make it run on dillon regularly,
instead of waiting until release day to possibly notice missing git
pushes (see bug report #762055).


KiBi.


signature.asc
Description: Digital signature