[RFC] eol values: unix/windows or LF/CRLF?

[Alexander Belchenko]

John Arbash Meinel пишет:
> 
> .....
> 
>>> Anyway, I'm thinking the actual values here don't really map at least to
>>> what *I* expected. Which is that if you set it to 'native', you would
>>> get local CRLF or LF on checkout, but always LF on checkin, and then
>>> things like 'unix', 'dos', and 'exact' would change what was in the
>>> working tree as well as what was in the repository. Given your formats:
>>>
>>> 'exact': []
>>> 'unix': [...(_to_lf_converter, _to_lf_converter)]
>>> 'dos': [...(_to_crlf_converter, _to_crlf_converter)]
>> ^-- is not this should be actually:
> 
>>  'dos': [...(_to_lf_converter, _to_crlf_converter)]
> 
>> I'm confused a lot. What the sense of having filter pair crlf/crlf?
> 
> Because you want to have crlf in the repository, and you want crlf in
> the working tree, but you know that a given editor/platform tends to
> screw it up.
> 
> bzr doesn't care what form is stored in the repo, as long as you are
> consistent it will give good compression, etc. The idea is to just
> enforce this fact at both commit and checkout.

OK, I undersrtand now.

> 
> John
> =:->
> 
>>> 'native': [...(_to_lf_converter, _native_outptut)]
>>>
>>> etc
> 
>