[darcs-devel] [patch2118] ci: include pending files in the git commit (and 21 more)
Ganesh Sittampalam
bugs at darcs.net
Sat Nov 21 10:38:46 UTC 2020
Ganesh Sittampalam <ganesh at earth.li> added the comment:
I confirm the tests pass on my Windows machine with this bundle. It's
fine to screen.
Doing a full review now of everything except the sendmail/MAPI patch
which I want to take a slightly closer look at.
> * ci: include pending files in the git commit
It might help to clarify in trigger-ci that it is testing the full
unrecorded state, not just the recorded state. Given that, using
--pending is natural (though the "darcs show files" docs suggest it
should be the default anyway?)
> * tests: more of the usual test repo removal fixes
OK
> * tests: simplify comparing patch bundles
Nice.
> * tests: compare paths in a more portable fashion
OK
> * fix broken tests/issue1101.sh
OK
> * ci: re-configure git globally on windows
OK
> * remove unused function resendEmail
OK
> * clean up some of the email sending mess
TODO
> * add a debug message to execSendmail
OK
> * more Windows test fixes
OK
> * fix tests/send-external.sh so that it works on windows, too
>
> Unfortunately it is not fully clear why invoking the generated script via
> bash does not work on Windows, whereas invoking a Haskell script via runghc
> works.
OK - I don't think invoking external scripts via bash has ever worked on
Windows, which is why there are typically Haskell helpers for the job. I
can't remember exactly why though and I agree in principle it seems like
it ought to work as people on Windows running the tests do have bash.
> * tests/add_permissions.sh: enable commented parts of the test
OK
> patch a84c0be7b93858fa34e66a50bea6662023a6ed75
> Author: Ben Franksen <ben.franksen at online.de>
> Date: Sun Nov 15 23:20:29 CET 2020
> * tests/lib: include \r in IFS on Windows
>
> This makes several of the tests succeed when I run the CI.
> + git config --global core.autocrlf input
Did you intend to include this line there too? If so I think it could do
with a comment as it's a bit weird for our core code/tests to be
invoking git.
> * tests/lib: factor out os_is_windows function
OK
> * ci: use ghc-8.8.4, not ghc-8.8.2
OK
> * ci: remove --hidesuccesses when running the tests
>
> It turned out that when we accidentally invoke MAPI on Windows, the command
> may hang indefinitely. With --hidesuccesses it is hard to detect that
> situation since running the tests on Windows is very slow.
OK
> * ci: add ACTIONS_ALLOW_UNSECURE_COMMANDS
OK
> * ci: use runner.os instead of fiddling with matrix.os
OK
>
> * ci: restructure the caching apparatus
OK
> * tests: remove obsolete setting of DARCS_EDITOR
OK
> * tests/printer.sh: disable test for \r encoding on Windows
>
> This fails for reasons that are still unclear to me.
The comment could more accurately say "This doesn't seem to work on
Windows in CI" as it does work for me, but not a big deal.
> * ci: enable running the tests on Windows
OK
__________________________________
Darcs bug tracker <bugs at darcs.net>
<http://bugs.darcs.net/patch2118>
__________________________________
More information about the darcs-devel
mailing list