Re: [PATCH v2] canonicalize-lgpl: Canonicalize casing too for MinGW.

[Paul Eggert]

On 12/11/21 03:35, Jan Nieuwenhuizen wrote:
> It looks like the inode returned by
> stat is always 0 on MinGW (at least it is under wine), so also when
> choosing this approach we would need to special case our client code for
> MinGW?

Yes, you'll have problems on such platforms. For ideas about this, 
please see the last several lines of:

https://www.gnu.org/software/gnulib/manual/html_node/sys_002fstat_002eh.html

(in particular the "partial workaround").

Since you're doing all this merely as an optimization, you can avoid 
doing the optimization on nonstandard platforms where stat is 
unreliable; that way, your code should still work even on nonstandard 
platforms, while having the optimized performance on standard platforms.