I setup some git hooks to run some gulp
commands on pre-commit. I basically run jshint
/plato
. I basically want to bypass these for two cases:
The plato gulp command runs analysis on the source and produces a /reports/ directory that tracks complexity over time. If we do this on the hotfix branch it will result in merge conflicts when merging them back into development. Enough talking here is the simple hook:
#!/bin/sh
if git diff --cached --name-only --diff-filter=ACM | grep '.js$' >/dev/null 2>&1
then
git stash -q --keep-index
./node_modules/.bin/gulp jshint
RESULT=$?
git stash pop -q
[ $RESULT -ne 0 ] && exit 1
git stash -q --keep-index
./node_modules/.bin/gulp plato
git add report/
git stash pop -q
fi
exit 0
Issue right now is if i have a merge conflict on "reports" and I resolve the merge All conflicts fixed but you are still merging.
and then commit it runs the analysis again and stages the commit and when it commits it throws an error:
/Users/Nix/work/project/.git/modules/somesubmodule/MERGE_HEAD' for reading: No such file or directory.
The directory does exist but there is no merge head...
So I just found a command that I think i can use to detect the "merge_head"
git rev-parse -q --verify MERGE_HEAD
If rev-parse returns a hash that means we are currently in a merge state. I can use that to bypass this logic. But will wait for some better advice from more experienced individuals.