When checko fails to push results for one repo, the whole pushing process stops
albertolluch opened this issue · comments
When checko fails to push results for one repo, the whole pushing process stops.
An example is when the server is not happy with the many connections and one gets a (kex_exchange_identification: read: Connection reset by peer). The manual trick, is to remove the groups that have been processed from the toml file, and try again.
Anther example, is if the group has no results.
It would be nice to skip the group and move on with the rest.
Should be fixed in 589d15d but it's a bit hard to test! Please reopen if the it still seems to crash!
I tried commit 206e7ad
I still got some blocking errors of 3 kinds:
(1) Groups without results
(2) Groups that were already pushed
(3) Connection issues.
Details below.
(1) Groups without results
To gitlab.gbar.dtu.dk:2023-02141/02141-group-2-eac46.git
32b6aff..daf7a41 results -> results
branch 'results' set up to track 'origin/results'.
Error:
0: could not read latest run at "submissions/e3048/run.json"
1: No such file or directory (os error 2)
(2) Groups that were already pushed
nothing to commit, working tree clean
Error:
0: command exited with non-zero code `git commit -m "Ran tests at Instant { t: 15454134367638 }"`: 1
Location:
checko/src/main.rs:187
(3) Connection issues
$ git push --force --set-upstream origin results
kex_exchange_identification: read: Connection reset by peer
fatal: Could not read from remote repository.
Please make sure you have the correct access rights
and the repository exists.
ERROR Group{name="1697c"}: failed to push results error="command exited with non-zero code `git push --force --set-upstream origin results`: 128"
INFO Group{name="9779f"}: cloning repo repo="git@gitlab.gbar.dtu.dk:2023-02141/02141-group-11-9779f.git"
Error:
0: An IO error occurred when talking to the server
1: kex_exchange_identification: Connection closed by remote host
I pushed a fix for this, which isolates all errors to a single group. Please close this when we know that the errors are not fatal!
In 1fcd27a I tried to improve this further, so hopefully it should be fixed by now! Please reopen this if you spot the error again!