[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: please ignore autopkgtest regressions in src:cub and src:libthrust



Hi Andreas,

On 08-03-2022 03:02, Andreas Beckmann wrote:
On 03-03-2022 11:28, Andreas Beckmann wrote:
with the recent uploads of src:cub and src:libthrust I've added autopkgtests, which unfortunately are causing regressions. Please ignore these for now while I'll try to fine-tune them in the future.

Are you in a hurry to get these versions in testing? If so, please

Currently the libthrust-dev dependency on libcub-dev is insufficiently versioned (only >= in testing), allowing mismatching upstream versions to be installed together, while the headers reject this combination (#1006226). Fixed in sid (using >= and <=), while testing has the broken combination:
libcub-dev    | 1.15.0-1      | testing    | all
libthrust-dev | 1.14.0-1      | testing    | all

I've hinted libthrust along to fix the testing situation. Why do we need to hurry cub?

elaborate why. Otherwise, I suggest to either just work on fixing the tests or disable them for now. Failing autopkgtests in testing are considered RC, so without good reason, we don't want to ignore regressions.

These are not regressions, only exposing problems already present in testing (but never noticed because there were no tests).

From our point of view it's a regression, albeit I know not everybody agrees with it. We go from no-failure to failure.

There are actually more issues than I initially expected, e.g. an incompatibility of nvcc with a header from g++-11 11.2.x, looks like a regression from g++-11 11.1.x. (astra-toolbox FTBFS #1003037)

Great that you found those. That sounds like progress.

Paul

Attachment: OpenPGP_signature
Description: OpenPGP digital signature


Reply to: