New patchwork state: Invalid

[Carlos O'Donell]

On Mon, Sep 19, 2022 at 05:49:30PM -0400, Siddhesh Poyarekar wrote:
> Adhemerval, Florian and I noticed a couple of manpage patches during review
> that managed to apply to the glibc tree because they only added new files
> and as a result, didn't fail CI.  I've added a new terminal state 'Invalid'
> to patchwork to mark such patches from now on.  It should be used for any
> patches that don't belong to glibc, e.g. because they've been cross-posted.

I noticed this too! I asked DJ aobut his and he said it was common.

I'm OK with "Invalid" but I would rather go with "Not applicable"
because the patch is *valid* but not applicable. Total bikeshed.

> This is now separate from the 'Fails to apply' state, which was previously
> overloaded to indicate non-glibc patches.

Our maxim should be: As few states as possible.

My opinion is that we should not have "Fails to apply."

Instead the state should always be "Failed CI."

Applying a patch is a lower level operation that is related to
pre-commit CI.

The flow could be like this:
- A patchwork bot attempts to re-apply the patch, and if it fails it
  creates a "fail" entry in the CI which reads "Rebase bot failed".
- A patchwork bot notices the fail and marks the patch or whole
  series as "Failed CI"

My suggestion is as follows:

- Remove "Fails to apply." from the workflow. As few states as possible
  should be our goal.
- Suggest renaming "Invalid" to "Not applicable" only because it sounds
  kinder (complete bikeshed).

Cheers,
Carlos.