krb5_cc_gen_new

[Jacques A. Vidrine]

On Fri, Mar 26, 2004 at 12:34:00PM -0500, Sam Hartman wrote:
> >>>>> "Jacques" == Jacques A Vidrine <nectar at FreeBSD.org> writes:
> 
>     Jacques> You can see from here that Heimdal's krb5_cc_gen_new, at
>     Jacques> least, is used *in place of* krb5_cc_resolve.
> 
> And has a different signature than our gen_new.  I cannot change the
> signature of an exported function even if it is completely broken.

Oh, I understand.  Was just trying to provide some example,
since there do not seem to be many consumers of this API.  If this
functionality is needed (I think it is), then perhaps krb5_cc_gen_new
should be deprecated, a new better-defined API introduced.

Or maybe Heimdal should just follow MIT's krb5_cc_resolve behavior,
which (IIRC) is to create a `randomly named' cache if the given name
is empty (e.g. "MEMORY:" or "FILE:").

> As a side note, if you could manage to call it, I don't think the use
> of mktemp in krb5_fcc_generate_new is unsafe on Unix systems.  IT can
> be DOSed, but should never overwrite files.

I admit that I haven't looked at the code to check.  As long as
something such as O_EXCL is used along with tight permissions, it
should be fine.

Cheers,
-- 
Jacques Vidrine / nectar at celabo.org / jvidrine at verio.net / nectar at freebsd.org