Hi,As usual, smcv expresses my thoughts much better than I can, but I'd like to comment on two points.
On 16-08-2026 11:18, Andrey Rakhmatullin wrote:
It gives the impression that deliverying as many packages aspossible is considered important.This is definitely a misfeature.
https://release.debian.org/testing/rc_policy.txt section 4 says: """Packages must be supported on as many architectures as is reasonably possible.""" The crux here being *reasonably*. There is no golden rule for this. We expect maintainers to judge in the spirit of Debian that carries more architectures than most upstream (officially) support. That's part of being Debian.
Anyway, i386 autopkgtest failures in Sid eventually kick out the package in Testing from all architectures.
I don't exactly understand what you mean here as failures in unstable don't kick packages out of testing. But some of the choices in our automation are trade-offs, so if a package might be blocked by such a choice please talk to the Release Team. An example that I can quickly think of where we're willing to add a hint [1] is, if a test of a package that isn't in testing fails on a particular architecture [2] while the source builds an arch:all binary (at least one) and it fails because of missing binaries on that architecture.
Paul[1] only needed one time as future uploads won't be considered regressions (unless a package is removed from testing again)
[2] except on amd64 and in principle also arm64
OpenPGP_signature.asc
Description: OpenPGP digital signature
