Mercurial > hg > hg-git
view TODO.txt @ 1047:cf982a23e15c
gitdirstate: show pattern error in hgignore file as expected
Before this revision, invalid pattern in hgignore file causes
unintentional failure for UnboundLocalError of ignorefunc, if hggit is
used with Mercurial 3.5 or later.
In such case:
- checking source of invalid pattern at failure uses "pats" list for
hgignore files, but
- "pats" list is empty, if ignoremod is None (= Mercurial 3.5 or later)
- therefore, checking with matchmod.match() overlooks invalid pattern
Then, "return ignorefunc" is executed without assignment to
ignorefunc, and causes UnboundLocalError.
To show pattern error in hgignore file as expected even with Mercurial
3.5 or later, this revision puts '(FILE, ["include: FILE"])' tuples
into "pats" (to avoid code duplication, putting into allpats is
shared, too).
This makes checking source of invalid pattern at failure work as
expected for hgignore files.
Fixes #197
author | FUJIWARA Katsunori <foozy@lares.dti.ne.jp> |
---|---|
date | Sat, 05 Aug 2017 02:13:11 +0900 |
parents | b826908ec522 |
children |
line wrap: on
line source
GENERAL ========== * work fine with eclipse plugin or tortoise-hg MAPPING ISSUES ============== * work in Git on a named branch created in Hg is forward-ported to be named branch commits in Hg and stripped back out if re-exported REMOTE/BRANCH STUFF ===================== * explain what branch mapping policy determined when updating refs * error nicer if pushing to remote without push access (over git://) WEBSITE =========== * more usage documentation * screencast SPEED/EFFICIENCY ================ * dulwich improvements - don't send blobs/trees already on server - thin packs - packfile creation benchmarking (seems to take a while sometimes) - at least provide status output MAYBE ========== * submodules? * .gitignore, etc - try to convert? - (probably not automatically, but perhaps a generator?)