Moving or fixing the .hgignore file is the right answer.

still an issue with bb9852b3bf06 (much after 0.9.1)

this is still an issue with Mercurial 0.9

I think this issue isn't yet in the 'testing' stage.

I think the proper solution would be to throw some sort of MatcherError
exception and catch it in the .hgignore case and not in the command line case.
By the way, Matt suggested it, the matcher functions could be implemented as a
class.
Sadly, I won't have time to work on this issue this month (i have a lot of work
for the end of the semester).

one thought that can maybe be safe: if error parsing ignore file, ignore
*everything*.

this way, every file not explicitly listed on command line will be ignored. this
will let you revert .hgignore by using "hg revert .hgignore", but e.g. "hg
revert some-dir-name" will do nothing.

i think this needs a patch to implement it and some testing to be sure it is safe.

here is patch to warn on error in hgignore file, instead of abort.

i am not sure it is right thing to do, but how else to undo mistaken edit of
hgignore?

[ is broken for sure.  if your shell finds a problem in a pattern, it usually
uses the text of the pattern and does not force you to fix the problem.

@vadim:
the '[' glob pattern is not a broken pattern, it is valid (it does the same
thing as a shell will do: match '[')

not sure how the '[' pattern managed to get past matcher, but I did find another
related bug (posted).

i see a problem. this hgignore file is broken, but mercurial does not report
anything:

syntax: glob
[

As of writing, Hg aborts on invalid .hgingore patterns, which is the right thing
to me. Perhaps a note can be added to doc on handling of bad patterns in .hgignore.

i think we need to be consistent:
- abort in case of error in .hgignore
- skip invalid patterns

actually we sometimes do the first option and sometimes the second.

I think this should be fixed now?

I don't like the disregard .hgignore option.

IHMO, 2 and 3 may have their own share of concerns, while 1 may be a good trick
that should go into Wiki until a solution is implemented. Now if I may...

4) add a global option to disregard .hgignore.

This is just an extension on 3, but allows user to decide this is what they want
to do.

tksoh, can you please test the attached patch, it implements option 2) (if
parsing of .hgignore failed, it will not use it).

I am not sure what to do with it.
1) We don't change mercurial, the user remove .hgignore and revert it. (nothing
to do)
2) We report invalid patterns and continue parsing .hgignore (harder to do).
3) We report invalid patterns and we continue without ignoring something (easy
to do).

If .hgignore contains invalid patterns, 'hg revert' aborts before reverting
anything - including the broken .hgignore itself. Apparently, it breaks 'hg
update -C' too.

The only workaround I found so far is to do 'hg cat .hgignore > .hgignore'