Interested in Participating

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Kent Gibson wrote:
> Hi John,
>> Sorry this took so long. Here are a few possible bugs that you might
>> want to look at. I can certainly give you more detailed pointers if you
>> are interested in any one of these. Most of them I have ideas about
>> possible fixes, just haven't gotten around to it.
>>
>> These have to do with how bzr ignores files:
>>
>> https://launchpad.net/products/bzr/+bug/3780
>> https://launchpad.net/products/bzr/+bug/4559
>> https://launchpad.net/products/bzr/+bug/57637
>> https://launchpad.net/products/bzr/+bug/29488
>>
>>   
> 4559 and 29488 are now fixed (well they both have patches in review
> anyway).
> So is anyone doing 3780 and 57637?
> Can you give me the more detailed pointers you mention?
> 

I'm CC'ing the mailing list in case anyone is interested.

I think Richard Wilbur is working on 3780. So make sure you assign 57637
to yourself so that you can avoid conflicts.

As for pointers for 57637:

You can dig up Jan Hudec's 'bzr.ignore' branch. His change is not a
trivial one that could be merged for bzr-0.12, but might be nice to get
into bzr-0.13.

The launchpad entry for Jan's branch is here:
https://launchpad.net/people/bulb/+branch/bzr/ignore

For a simpler short-term fix which might make it into 0.12, you could
look at the code for fnmatch.translate(), and change it so that the glob
'*' isn't expanded to the regex '.*' but instead something like '[^/]*'.


> In a similar vein, as part of Aaron's review of my 4559 changes where I
> now reject absolute paths, he suggested we may want to accept absolute
> paths - so long as they terminate in a branch (as per bzr add).  Does
> this sound reasonable to you as well?  Aaron is concerned it may be
> controversial.
> 
> Cheers,
> Kent.
> 

I think it is reasonable for paths to go through something like
"WorkingTree.relpath", since that changes both relative and absolute
paths to be relative to the tree root.

Then it doesn't matter if you are in a sub directory, or are passing a
path for another directory 'bzr ignore ../other/file/in/project', etc.

The only controversial bit is what to do if given:

cd ..; bzr ignore project1/foo project2/foo

'bzr add' fails with an error, which is reasonable, but it shouldn't
show a traceback. It should fail with something like:

bzr: ERROR: Path 'project2/foo' is not in 'project1'.

Probably getting nice short paths is difficult, so something like:

bzr: ERROR: Path '/home/user/project2/foo' is not a child of
'/home/user/project1'

So a similar error message, just without the whole traceback. (Which
could be done with is_user_error=False, but PathNotChild may be used in
other ways, so an audit of usage should be done to figure out if a
different exception should be used).

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFNcdwJdeBCYSNAAMRAvopAJ9ARhiAD2GsBqqhzj0tXzHzPuAArgCfecMd
32jchj1YGJ1QabY8Ojnl+JI=
=Likm
-----END PGP SIGNATURE-----