Mercurial case-folding BIG BUG

[Filippo Pedron]

Dustin Sallings wrote:
>
> On Feb 3, 2008, at 14:14, Filippo Pedron wrote:
>
>>>     If you disable the check, then what happens when the collision 
>>> occurs?
>>>
>> I understand that this is not the solution but only an emergency 
>> solution.
>
>     I don't see how it makes anything better at all.
>
>>>     If your filesystem can't tell the difference between FOO.TXT and 
>>> foo.txt, how should mercurial (or any revision control system) apply 
>>> changes to one or the other (or Foo.Txt)?
>>>
>> I agree with you and I understand that my example was too much simple 
>> to explain the problem. Follow this other (reds are added/modified 
>> lines):
>
>     [stuff]
>
>> > ren foo.txt FOO.TXT
>> > hg add FOO.TXT
>
>     Does ``ren'' mean rename?  If so, that can't possibly be what you 
> intended to do.
>
>     From mercurial's perspective, all you did was add (see the ``hg 
> add'') a new file with the same name as another one, but in a 
> different case.
>
>     Perhaps you wanted to ``hg mv'' or at the very least ``hg rm -A 
> foo.txt''
>
>     What you are doing here is specifically configuring the tool to 
> attempt to do something your filesystem cannot handle.  Mercurial now 
> sees a foo.txt and a FOO.txt.  Your OS can't tell the difference.  
> Mercurial is trying to tell you that.  You should listen to it (and 
> don't just go changing mercurial to ignore important warnings).
>
"ren" means "rename" and I used it just to show what happen.
In my project, a friend of my sent me a renamed file and for Windows 
everything was ok. I added some other file and committed a lot of time. 
No warning on the command line, no warning in the manual, hg tests were ok.
After a while, I needed to revert changes and I wasn't able to do it.

I think that on my Windows system, Mercurial should abort to commit if 
it find a case-folding collision. In this case, I should do something as 
you suggest (rename or move or remove the file) to solve the collision 
or "force" hg to commit with some special flag.

In the update case on Windows system, hg should work as some ftp client 
( FileZilla, ... ): hg should ask something like [skip, overwrite or 
abort] in some interactive way. It should also give a warning feedback 
as final message. Finally,  to solve case-folding collision problem, I 
should be able to call hg to revert a file if it match exactly the case.

On Linux everything works smoothly because there isn't a case-folding 
collision problem.

Best regards,

Filippo Pedron



 
 
 --
 Email.it, the professional e-mail, gratis per te: http://www.email.it/f
 
 Sponsor:
 Ti vestiamo dalla testa ai piedi e anche sotto se lo vuoi ! Clicca su www.grandinettisport.com
* 
 Clicca qui: http://adv.email.it/cgi-bin/foclick.cgi?mid=7502&d=4-2