Re: Re-casing directories on case-insensitive systems

From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Jan 11, 2008, 21:18 UTC
Message-ID: <alpine.LFD.1.00.0801111313220.3148@woody.linux-foundation.org>
In-Reply-To: <579DF776-4F4E-464C-88DB-B22C2EC291BD@sb.org>

On Fri, 11 Jan 2008, Kevin Ballard wrote:

Show 8 quoted lines

>
> Anyway, here's the state of my directory:
> 
> kevin@KBALLARD:~/Documents/School/C07> git ls-tree HEAD
> 040000 tree b47c8103e2e01fcf145bdc237c0e56ffc61f1c47	CS4536
> 040000 tree dbf7fc51ef3effebdf9b4e9172e4c86cae52b163	cs4536
> 040000 tree 15834a7b6534a285bf6930be4e5404b37e1dc718	ece3601
> 040000 tree 62d229b8c4a389b550df20a3752d666c48c767a4	ma2071

Hmm. You can do something like

	git ls-files CS4536 | xargs git update-index --force-remove

which will remove gits knowledge of that directory even though "lstat()" will still claim that all the files still exists.

Case-insensitive filesystems are a pain.

I wish we had some way to handle it sanely, but I don't think a sane solution to case-insensitivity exists. If you limit it to strictly just 7bit ascii names (like in your example), then some of the problems do go away, but it would still be probably fairly major surgery to try to teach git about the whole insanity of a case-independent working tree.

		Linus