W. Martin Borgert wrote: > > Naturally, downloading them during the build would not be > > acceptable, but I wonder what the consensus would be on the ethics > > of downloading "contrib"-like data during autopkgtests? > > It's also a practical matter: Can I run autopkgtests, when my > computer is disconnected from the internet? Can I run them, if > their server is down? Will they always serve exactly the same > files or will I have to deal with random test results?
Indeed, this would definitely be a vastly inferior fallback option. As someone who has encountered these issues countless times via the Reproducible Builds effort, I might have assumed these highly-salient practical concerns as a given and focused on the free software question. (However, on the practical side, I believe the "flaky" Restriction could, potentially, be suitable.) > Maybe we don't need a huge amount of test data, i.e. not the > full set that is used by upstream. A small subset might be > sufficient for at least some packages. In my case, it's only > 3.3 MiB. Noted, but it is not the size that is the concern, but rather the licensing. Regards, -- ,''`. : :' : Chris Lamb `. `'` la...@debian.org 🍥 chris-lamb.co.uk `-