El 22/2/25 a las 18:20, Marco d'Itri escribió:
On Dec 27, Santiago Vila <sanv...@debian.org> wrote:
During a rebuild of all packages in unstable in the year 2028, your package
failed to build:
This is expected, due to its specific nature.
If it will be needed to rebuild the package over the lifetime of trixie
then the certificate will be updated.
Hmm, I'm not happy that we do nothing about this.
The idea was not to have time bombs in the distribution at all, i.e. stop
allowing packages with such "specific nature".
Having a time bomb and saying "yes, there is a time bomb, but don't worry
because somebody will deactivate it before it explodes" does not help to achieve
such goal. Think about the "desert island test" which we use to evaluate
DFSG-freeness, but for building packages.
It is not possible to stop running the test when we know that it will fail?
I mean something like this, in pseudocode:
if currentdate < expirationdate:
run-test-which-we-know-will-fail-after-expiration-date
If we know the exact date in which it fails, letting the test to fail
after such date does not convey any useful information about the good/bad
state of the package.
Also: What about moving the verification part to dh_auto_test. so that at least
it will work when using nocheck?
Thanks.
