<task/<NAME>/failure> should also set error code in condition

[hongkongkiwi]

It would be great if when a task fails, that we can set a condition based on the failure code. This way specific scripts can be done to deal with the results.

For example, lets say we have a script which checks whether a hardware has failed. It can return 0 (success) or (failure) with exit code 1,2,3 so it would be great if for example it failed with error code 1 then it set both
<task/<NAME>/failure> and <task/<NAME>/failure:1>
error code 2 would look like:
<task/<NAME>/failure> and <task/<NAME>/failure:2>
etc

This would allow specific scripts to run if you know in advance what the error code might be. If you don't know, you can still using the existing <task/<NAME>/failure> condition.

[hongkongkiwi]

clearing conditions then might be slightly more tricky than currently, you would have to do (if in bash):
rm -f "/run/finit/cond/task/<NAME>/failure" && rm -f "/run/finit/cond/task/<NAME>/failure:"*

[troglobit]

No, that's not what the conditions system is supposed to do. Not to mention the unnecessary complexity it would add.

I'd much rather see we handle this as part of #273 instead.

[hongkongkiwi]

So your thinking is that with the above fix, there can be a catchall script run upon failure which then checks the errorcode using initctl -s status and takes action that way?

[troglobit]

Yes exactly, 1) task failed, start script, 2) check exit status (and/or state) of task. This moves a lot of the logic to initctl, instead of squeezing it into PID 1.

Does that make sense for your use-case?

[hongkongkiwi]

Yep, that would work fine thanks!

[troglobit]

The post: script action and the just pushed f4d3cf2 should be sufficient for this issue. See #273 for more information.