git-confirm
git-confirm copied to clipboard
pimterry
git commit will hang if lots of files are in staging, like when merging master
for now I am using git commit --no-verify . but maybe git confirm can have some kind of logic to skip the check if there are too many files in staging,
That should still work fine, I've done commits of similar sizes before with no problems. How large are the files overall?
CHARACTERS 88114 WORDS 6335 SENTENCES 1258 PARAGRAPHS 1728 WHITESPACE 23160
i don;t want to put the git diff file as it is internal repo. but i could replace the characters in the file with some token and post it
Is that the size of the diff, or the files themselves?
Number of lines would be useful too.
Hmm, even that should be fine. What git-confirm actually does is:
-
For each pattern, for each changed file, get the diff:
changes=$(git diff -U999999999 -p --cached --color=always -- $filename) -
Grep that diff for added/changed lines:
echo $changes | grep -C4 $'^\e\\[32m\+.*'"$match_pattern" -
(and then report any matches)
The interesting question is which part of that is slow for you, and why. If we know that, we can work out how to detect this case quickly and reliably, and then do something about it (or stop it being slow in the first place).
If I take the diff you've uploaded there, and git add it in a fresh repo, doing this is pretty much instant, so that alone isn't the problem. I can't actually remember why we include the -U999999999 here, and that might get expensive with very large files only containing a few changes, since we're effectively grepping the whole thing. Could that be the problem?
There's also generally some duplication here, I don't think there's any good reason not to do the diff once per file and then grep repeatedly, rather than repeating everything for every pattern. Do you have a very long list of custom patterns configured?
ah no.
#!/bin/sh
. git-sh-setup # for die git-diff-index -p -M --cached HEAD -- | grep '^+' | grep @#rbc && die Blocking commit because string @#rbc detected in patch
so i can run git confirm sh seperately then i guess to see where the bottleneck is
could u give me an sh file which outputs the start time and end time after each operation , so then i can see where it is taking time ?
The real code is here: https://github.com/pimterry/git-confirm/blob/master/hook.sh#L50-L55
You can add date on line 50, 52 & 55 to get printed timestamps before & after each step.
If you've got your repo in the right state (all your changes added, ready to commit) you should also be able to just run the commands above with the appropriate $filename and $match_pattern to reproduce this.