Fix multiple comments on label change failure

[yadij]

Only add comments when the labels actually changed.
The GH tool still needs to be updated to restore proper
behaviour.

[rousskov]

AFAICT, this script incorrectly ignores other [gh] failures as well. I also sense a disconnect/mismatch between failing to remove a label and the subsequent PR creation attempts that ignore that and other failures. There are other inconsistencies/puzzles as well. Said that, the proposed fix is probably a step forward.

[rousskov]

The GH tool still needs to be updated to restore proper behaviour.

FWIW, I do not know what that comment in PR description means/implies.

[kinkie]

The GH tool still needs to be updated to restore proper behaviour.

FWIW, I do not know what that comment in PR description means/implies.

It's comments like squid-cache/squid#2314 (comment)

[rousskov]

The GH tool still needs to be updated to restore proper behaviour.

FWIW, I do not know what that comment in PR description means/implies.

It's comments like squid-cache/squid#2314 (comment)

AFAICT, you are giving an example of a GitHub comment that PR-affected code is producing. I understand what those GitHub comments are. I do not understand the specific PR description paragraph that I have quoted.

[yadij]

The GH tool still needs to be updated to restore proper behaviour.

FWIW, I do not know what that comment in PR description means/implies.

It's comments like squid-cache/squid#2314 (comment)

AFAICT, you are giving an example of a GitHub comment that PR-affected code is producing. I understand what those GitHub comments are. I do not understand the specific PR description paragraph that I have quoted.

For the non-error process to work as intended, the gh binary that these scripts runs needs to also be upgraded. Otherwise we will just continue to have label errors, but without the duplicate comments.

I have done that upgrade in the background, which is why the comments stopped happening already before this PR merge. As I write this the script error handling is still broken, but not triggered by the latest binary.