On Fri, Oct 16, 2020 at 01:22:32PM +0200, Bjorn Ketelaars wrote: > Diff below brings notmuch to 0.31. From NEWS [0]: > - notmuch now supports Emacs 27.1 > - add support for moving between threads after > notmuch-tree-from-search-thread > > The test suite completes with 4 failing tests, test.log attached: > > 1210/1219 tests passed. > 3 broken tests failed as expected. > 4 tests failed. > 2 tests skipped. > All tests in 1 file skipped. > > I'm not sure if these failures are new. Current version in ports hangs > during 'make test' after 'Message with quote in Message-Id:'. Probably > related with Emacs 27.1. Unfortunately I don't understand these failures > and could use some help. > > No issues found while using this update, in combination with neomutt, on > amd64. Python bindings have been build tested only. > > [0] https://notmuchmail.org/news/release-0.31/
Thanks for the diff, a few comments: - REVISION-main = 0 should be removed - the changelog mentions "Add notmuch_message_get_flag_st and notmuch_message_has_maildir_flag_st, and deprecate the existing non-status providing versions", and OpenBSD's src/lib/check_sym confirms that notmuch_message_get_flag was removed. I don't think that anything in the ports tree uses this symbol, but to be safe one should probably bump the library major number. - the "realpath" non-portability issue was recently fixed upstream, replacing "realpath emacs" by "cd emacs && pwd -P", so perhaps the same fix could be applied for 0.31 in ports. About the tests: - I don't think that I experienced a hang in the tests in previous versions. - T360-symbol-hiding is just a linking (with relative path) problem, I tried fixing it manually (not a permanent solution) and it passed. - The "Yen tests" were already failing before, and are not real failures (there is more than one codepoint representing the Yen symbol, if I remember correctly one of these is fullwidth). - Now the segfault in "notmuch_database_get_config_list: closed db" in T590-libconfig is worrying. I'll send a bug report to the notmuch mailing list right now. Note that this test is new in 0.31, so that might not be a new problem. Also it is unlikely to be triggered in real life (close a db and then try to access it) so I don't think that should be a blocker for the port.