On 2023/02/03 12:17, Marc Espie wrote: > On Fri, Feb 03, 2023 at 11:07:26AM +0000, Stuart Henderson wrote: > > On 2023/02/03 10:40, Marc Espie wrote: > > > This was prompted by a discussion to kn@ > > > I'm not sure the use case is that frequent to warrant > > > the patch, plus it will obviously trigger if we > > > change mirrors project-wide. > > > > > > but it's just a warning, so probably not a big issue. > > > > > > (we no longer have user-side changes to MASTER_SITES* ever > > > since it ended up in sqlports as well) > > > > > > > > > What do you guys think ? > > > > It does seem that it maybe a bit annoying when people have reordered > > network.conf to list local mirrors first (or added their own known good > > ones), especially if those people are committers. > > > > What's the use case it's intended to help with? > > > > > Apparently kn@ managed to tweak the MASTER_SITES on a port right before > committing and did not notice the distfiles provided by the new MASTER_SITES > were different. > > It's probably a bit of a fringe case.
Been there, done that (usually switching between pypi and a git archive for something that doesn't normally include tests on pypi). I think if I'd remembered to do checksum/makesum I'd probably have remembered to clean distfiles too. Maybe it helps in some cases but I'm a but dubious that it's worth the cost (~10k distinfo files). I've currently (thanks github ;) got "make checksum DISTDIR=$tempdir" in my pre-commit check script which is a better check, but that's a bit annoying too. > and yeah, I don't think we have enough introspection to handle that kind > of thing, like manual sources of distfiles that don't come from network.conf. >
