The network/log test is still occasionally failing in the CI, though never
when I run the tests on my computer. Re-running the failed tests usually
"fixes" the problem. The two patches that don't exclusively concern the CI
are included because they are the result of my attempts to debug this
problem.
6 patches for repository http://darcs.net/screened:
patch 61494de595725bea4ba1babb4499a5dcaa9a265e
Author: Ben Franksen <ben.franksen@online.de>
Date: Sat Nov 19 21:25:58 CET 2022
* ci: checkout and cache actions v2 -> v3
patch b6a14c2e6c0d0f6c969cd383fd610b610b7d8e67
Author: Ben Franksen <ben.franksen@online.de>
Date: Mon Nov 21 16:00:47 CET 2022
* support ghc-9.4
patch 620162ef5e1f96baf6e292aec6789d944fc7f57d
Author: Ben Franksen <ben.franksen@online.de>
Date: Tue Nov 22 08:48:17 CET 2022
* ci: reformat version matrix, improve cache hit rate
patch 5f10c28d1147d096a986da26da8e8721ec54d1ef
Author: Ben Franksen <ben.franksen@online.de>
Date: Tue Nov 22 09:40:11 CET 2022
* tests/network/log.sh: create a read-only dir instead of using root dir
patch d92b0ac32c52654a850094fb81f6b254379ed815
Author: Ben Franksen <ben.franksen@online.de>
Date: Tue Nov 22 19:17:25 CET 2022
* improve error reporting when remote _darcs/format doesn't exist
The subtlety here is that if we cannot fetch _darcs/format, we then try to
fetch _darcs/inventory, and if that succeeds we conclude that we have a
darcs-1 repo. However, if that fails, too, then we reported the exception
for _darcs/inventory and not the one for _darcs/format, which is completely
useless to find out what the problem is, unless the repo you want to access
is actually an ancient darcs-1 repo with no format file. So instead, we now
report both exceptions.
patch 23fc4bd93bd18dc29e5573ba37474a215f22b436
Author: Ben Franksen <ben.franksen@online.de>
Date: Tue Nov 22 09:45:45 CET 2022
* ci: update the available runners (OS versions)
As before, all but the latest versions are commented out.
Attachments
|
