On Mon, Jun 16, 2025 at 10:17 PM Branko Čibej <br...@apache.org> wrote:
> On 17. 6. 25 03:29, Nathan Hartman wrote: > > On Mon, Jun 16, 2025 at 4:23 PM Branko Čibej <br...@apache.org> wrote: > >> When building trunk (with autotools and maintainer-mode; to test with >> serf-2), I get a lot of warnings like this one: >> >> .../subversion/svnadmin/svnadmin.c:2598:11: warning: >> 'svn_opt_args_to_target_array3' is deprecated [-Wdeprecated-declarations] >> 2598 | SVN_ERR(svn_opt_args_to_target_array3(&targets, os, >> | ^ >> >> >> These are relatively recent. Policy says that trunk should compile >> without warnings in maintainer mode. This is on macOS with clang, there are >> similar warnings in the autoconf worflow on GitHub. >> > >> >> There used to be a warnings-check builder that would fail in such cases, >> but apparently it hasn't been ported to GH actions. >> >> -- Brane >> >> Hi Brane, > > Can you tell us a bit more about the policy that trunk should compile > without warnings in maintainer mode? I ask because it doesn't seem to be > documented in HACKING, and it seems surprising to me, since during > development I would hope to see maximum warnings to catch issues early. > (I'll be happy to improve HACKING once I have a better understanding...) > > > > It might not be documented at that ... but it's sort of implied in the > idea that trunk should be stable and ready to cut a release at more or less > any time. Obviously it's not a hard rule or we couldn't have trunk-based > development + CTR. And no-one was ever expected to test their changes on > more than one platform. > > -- Brane > Oh, I misinterpreted what you wrote before. I interpreted it as "builds in maintainer mode should hide warnings" but should have interpreted it as "warnings shouldn't happen because trunk shouldn't be broken." My bad! Cheers, Nathan
