Re: [PATCH 0/6] Extensions of core.ignorecase=true support

From: Johannes Sixt <j6t@kdbg.org>
Date: Aug 18, 2010, 18:31 UTC
Message-ID: <201008182031.10197.j6t@kdbg.org>
In-Reply-To: <AANLkTi=Fkj-_-wHZN4GY2b+2n0mk7MKV=au0Hm4B+i9d@mail.gmail.com>

On Mittwoch, 18. August 2010, Robert Buck wrote:

> So apparently when core.ignorecase=true, this really means
> casepreserving=false, casesensitive=true.

No. When an entry enters the index from the worktree, an existing entry is reused and the index entry's case is preserved, where existence is determined in a case-insensitive manner[*]; if the entry is new, the case is preserved.

In the oppsite direction, case preservation depends entirely on the capabilities of the file system.

[*] The new part of this series is that this case-insensitive existence test happens on for the entire path, and not just the file name part.

Show 5 quoted lines

> So what I am hearing is that unless one sets core.ignorecase, in mixed
> environments you are in for a world of hurt; you'd end up with Foo and
> foo from the Unix side of the house, and on Macs or Windows the last
> file materialized from the index or repository into the working
> directory would clobber the first one materialized,

core.ignorecase is not designed to help this case. It is a "Doctor, it hurts when I poke myself in the eye" problem. Don't have both foo and Foo in your repository if you go cross-platform. The git repository format is not designed to treat them as identical, and no configuration option alters this.

-- Hannes