You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Oct 11, 2023. It is now read-only.
Errors were found upon further inspection of the Google Cloud logs, therefore weakening the argument for paritytech/polkadot#2610 (comment) being in the right trail for that situation (it'd be part of another problem: #236).
This issue therefore lists findings about the current problems with the Companion Update Flow not related to how the build system currently manages concurrent merge requests, but instead centered around how failure is badly handled currently. While #236 and #250 can't be solved reliably in the short term, what we can do for the time being is post a comment when an error happens so that developers can resolve them manually.
a git clone is attempted. It might fail, and it indeed failed in that situation, although the error didn't matter. Either way, failure is not handled currently, but it should be. From the logs:
fatal: destination path 'polkadot' already exists and is not an empty directory
Merge might be done on a outdated master branch
The merge master command does not fetch origin/master before attempting the merge, although it should, since it's only guaranteed to be up-to-date on a fresh clone (which might fail, as described above).
{
"textPayload": "[2021-03-16T12:04:05Z INFO parity_processbot::companion] Merging master.\n",
},
{
"textPayload": "Already up to date.\n",
}
Regardless of whether or not origin/master was indeed up-to-date with Github at that point in time, this measure still needs to be taken into consideration.
Due to the bugs mentioned before, no change was made to the repository, which meant there was nothing to commit
"textPayload": "[2021-03-16T12:04:18Z INFO parity_processbot::companion] Committing changes.\n",
},
{
"textPayload": "On branch gav-abstract-payout-curve\n",
},
{
"textPayload": "Your branch is up to date with 'temp/gav-abstract-payout-curve'.\n",
},
Even if those commands were properly checked for their exit status, as it currently is the bot would still fail silently with a single uninformative comment (as in paritytech/polkadot#2610 (comment)) because those errors are simply logged without posting a comment
This issue is related to the scenario in paritytech/polkadot#2610
Errors were found upon further inspection of the Google Cloud logs, therefore weakening the argument for paritytech/polkadot#2610 (comment) being in the right trail for that situation (it'd be part of another problem: #236).
This issue therefore lists findings about the current problems with the Companion Update Flow not related to how the build system currently manages concurrent merge requests, but instead centered around how failure is badly handled currently. While #236 and #250 can't be solved reliably in the short term, what we can do for the time being is post a comment when an error happens so that developers can resolve them manually.
Problems
Cloning might fail
In
parity-processbot/src/companion.rs
Line 65 in dee4d24
git clone
is attempted. It might fail, and it indeed failed in that situation, although the error didn't matter. Either way, failure is not handled currently, but it should be. From the logs:Merge might be done on a outdated master branch
The merge master command does not fetch
origin/master
before attempting the merge, although it should, since it's only guaranteed to be up-to-date on a fresh clone (which might fail, as described above).parity-processbot/src/companion.rs
Line 120 in dee4d24
It showed up in the logs:
Regardless of whether or not
origin/master
was indeed up-to-date with Github at that point in time, this measure still needs to be taken into consideration.The bot fails silently
The logs shows that the cargo update command had failed:
Due to the bugs mentioned before, no change was made to the repository, which meant there was nothing to commit
Even if those commands were properly checked for their exit status, as it currently is the bot would still fail silently with a single uninformative comment (as in paritytech/polkadot#2610 (comment)) because those errors are simply logged without posting a comment
parity-processbot/src/webhook.rs
Lines 1281 to 1284 in dee4d24
Solution
The text was updated successfully, but these errors were encountered: