Hi,

Le mardi 02 août 2022 à 21:43 +0200, Sebastian Ramacher a écrit :
> On 2022-07-31 13:23:38 +0200, julien.pu...@gmail.com wrote:
> > Package: release.debian.org
> > 
> > Some Coq-related packages need a rebuild:
> > 
> > coq-hierarchy-builder
> > mathcomp-algebra-tactics mathcomp-analysis
> > 
> > where packages on the same line can be handled in parallel.
> > 
> > I can't give a nice ben script because the abi checksum varies with
> > the
> > architecture (see today's mail on debian-devel where I'm trying to
> > find
> > ideas for a better approach).
> 
> From the discussion on -devel, a permanent tracker like the one for
> Haskell (https://release.debian.org/transitions/html/haskell.html)
> could
> help with the rebuilds for coq-* and related packages. Do all
> affected
> packages depends on some package that we can use as a basis for the
> permapermanent tracker?

Short answer: no.

Longer answer: I have some code to manage the Coq-related packages,
which can answer simple questions and might answer complex questions.

For example, if I want to update src:coq-elpi, I can run:

./planif_transition.py coq-elpi
coq-elpi
coq-hierarchy-builder
mathcomp-algebra-tactics mathcomp-analysis

which tells me what packages I can update and in which order ; some can
be parallelized (same line).

Since the thread on debian-devel was started, I tried to write another
script, aptly named wanna-build, but I'm not sure I got everything
right. If I wanted to upload the new upstream of coq-elpi, the plan
would be:

./wanna-build.py coq-elpi 1.15.5-1
 nmu coq-hierarchy-builder_1.3.0-1 . ANY . -m 'Rebuild due to new coq-
elpi 1.15.5-1'
 dw coq-hierarchy-builder_1.3.0-1 . ANY . -m 'coq-elpi => 1.15.5-1'
 nmu mathcomp-algebra-tactics_1.0.0-6+b1 . ANY . -m 'Rebuild due to new
coq-elpi 1.15.5-1'
 dw mathcomp-algebra-tactics_1.0.0-6+b1 . ANY . -m 'coq-elpi => 1.15.5-
1'
 nmu mathcomp-analysis_0.5.2-2 . ANY . -m 'Rebuild due to new coq-elpi
1.15.5-1'
 dw mathcomp-analysis_0.5.2-2 . ANY . -m 'coq-elpi => 1.15.5-1'
 dw mathcomp-analysis_0.5.2-2 . ANY . -m 'coq-hierarchy-builder =>
1.3.0-1+b1'

You might be worried that the planif-transition and wanna-build scripts
don't agree on the right time to build mathcomp-algebra-tactics. The
reason is that the first looks at the whole dependency graph of Coq-
related packages while the second only sees the extracted affected
packages dependency graph: both are right, and guarantee a sane
building order.

Does that look sane?

Cheers,

J.Puydt

Reply via email to