Junio C Hamano, Wed, Oct 27, 2021 22:16:35 +0200:
Alex Riesen [off-list ref] writes:
quoted
From: Alex Riesen <redacted>
This allows re-enabling hooks disabled by an earlier "--no-verify"
in command-line and makes the interface more consistent.
Signed-off-by: Alex Riesen <redacted>
---
This one is on top of "[PATCH] Fix "commit-msg" hook unexpectedly called for
"git pull --no-verify" (http://public-inbox.org/git/YXfwanz3MynCLDmn@pflmari/).
Which is a bit awkward. Should I resend as series?
Don't we need to do this at the root cause command "git commit"?
The commit preparing code in builtin/merge.c does not seem to use the code
from builtin/commit.c, so it does not look like a direct cause of that effect
in "git merge". But...
It is documented to take "--no-verify" but not "--verify" to countermand an
earlier "--no-verify" on the command line.
This particular peculiarity in implementation of "--[no-]verify" does look
like it has root-caused everything :)
And yes, I agree that we shouldn't introduce an awkwardness in one
step of the series and fix it in another step of the same series.
I think I better resend everything as a single patch then.