On Thu, Jan 18, 2001 at 11:11:13AM +0100, Steffen Neumann wrote:
> 
> Hi,
> 
> When applying the pdbtool import function
> to a stock Unix group file, things will break 
> because standard gids are (to my knowledge) positive.

Yeah, the import/export use a format similar to /etc/passwd and
/etc/group just because it is a more familiar format. Never thought
about importing real groups.

The main differences are that Coda groups are identified by negative
numbers. And they have the concept of an `owner', a user who would
normally be allowed to add and remove group members. It is just that
`au/auth2' don't have that capability at the moment.

The owner is `encoded' as the first group member in the exported group
file.

> Is there any use for our patch ? 
> Or did we break something else ?

It is probably useful. Although the group ownership is something that
could give unexpected results.

> 607c607,609
> <       rc = fscanf(userfile, "%[^:]:%*[^:]:%d:%*s\n", user, &user_id);
> ---
> >       /* %s SEEMS BUGGY ! STN */
> >       /* rc = fscanf(userfile, "%[^:]:%*[^:]:%d:%*s\n", user, &user_id); */
> >       rc = fscanf(userfile, "%[^:]:%*[^:]:%d:%[^\n]\n", user, &user_id);

There is no third argument, so it would at least be ":%*[^\n]\n" to
avoid fscanf assigning to an argument that was never passed.

Jan