[SCM] Samba Shared Repository - branch v3-2-test updated - release-3-2-0pre2-2967-g0cba300

[Jeremy Allison]

On Fri, Aug 29, 2008 at 10:58:39AM +0900, SATOH Fumiyasu wrote:
> At Thu, 28 Aug 2008 14:11:15 -0500 (CDT),
> Jeremy Allison wrote:
> > commit 0cba30073410ba499834ac26dee3f81a75c8de4f
> > Author: Jeremy Allison <jra at samba.org>
> > Date:   Thu Aug 28 12:09:06 2008 -0700
> > 
> >     Add st_birthtime and friends for accurate create times on systems that support it (*BSD and MacOSX). This really needs to be in 3.2.x.
> >     Should have done this ages ago, sorry.
> >     Jeremy.
> 
> If a filesystem (e.g. UFSv1) does not support birthtimes,
> pst->st_birthtime* have invalid value.
> You should ignore pst->st_birthtime* on that case
> and fallback to legacy code (calc_create_time()).
> 
> How to check if pst->st_birthtime* are valid or not?
> I think Samba should fallback to calc_create_time()
> when:
> 
>   (1) pst->st_birthtime == (time_t)0 (if pst has no st_birthtimensec)
>   (2) pst->st_birthtime == (time_t)0 and pst->st_birthtimensec == 0L
>   (3) pst->st_birthtime == (time_t)0 and pst->st_birthtimensec == -1L
>   (4) pst->st_birthtime == (time_t)-1 (if pst has no st_birthtimensec)
>   (5) pst->st_birthtime == (time_t)-1 and pst->st_birthtimensec == -1L
> 
> I've NOT confirmed that there is standard specification for
> st_birthtime* (especially, what value indicates that
> a filesystem does not suport it), and the above (1)-(5)
> conditions are suitable for judgement.
> 
> FYI.
>   http://www.nabble.com/birthtime-initialization-td18584558.html

Hmmmm. So what you're saying is that they added it, but completely
f*cked it up so it's not at all useful and you can't even tell if
it's valid or not...

Great. *BSD innovation at it's finest. Almost as bad as Linux aio :-).

Ok, I'll add this tomorrow when I'm back in at work.

Jeremy.