bzr (linux) doesn't handle FAT's limitations (X-bit and symlinks) well. Should behave like on windows.

[alsuren]

Steps to reproduce:

create a repo on an ext filesystem, and do some stuff
copy it to a fat32 memory stick
watch bzr status think that everything's changed permissions.

add a symlink on the ext side, and do it again
watch bzr give a really unhelpful permission denied error.

This is what the chaps in #bzr think about the issue:

(11:49:27) lifeless: so the basic problem is, fat is returning garbage data, and we assume when the platform is unix that all fs's track X reasonably well
(11:49:39) lifeless: we have code to work around no X tracking but only activate it on windows
(11:49:42) lifeless: file a bug please
(11:50:04) lifeless: I suggest we add a a config setting/environment variable to force that workaround code on in this sort of situation
(11:50:43) Peng_: What about explicitly checking if the fs supports it?
(11:50:51) Peng_: Same way you do the CIFS checks?
(11:51:13) lifeless: Peng_: what CIFS checks
(11:51:23) lifeless: Peng_: do you mean the case insensitive ones?
(11:51:28) Peng_: Yeah.
(11:51:37) lifeless: we need either reference data or to write
(11:51:48) lifeless: its possible in principle

[alsuren (alsuren)]

There are a few OS level bugs that may be related to this one:, eg.

https://bugs.launchpad.net/ubuntu/+source/gnome-mount/+bug/178154

but checking whether your working tree supports chmod et al is still useful. See also:

https://bugs.launchpad.net/bzr/+bug/190725

[alsuren (alsuren)]

see also https://bugs.launchpad.net/bzr/+bug/190725 and https://bugs.launchpad.net/ubuntu/+source/gnome-mount/+bug/178154

The second link contains the workaround I'm currently using for the permissions (I decided that setting the permissions to 644 was better than all 755 for my project.)

[Robert Collins (lifeless)]

Working with mounting FAT fs's isn't a core use case at the moment. It is something we can improve on though.