libbsd

[Rusty Russell]

On Sat, 30 Jun 2012 19:38:32 +1000, Andrew Bartlett <abartlet at samba.org> wrote:
> On Fri, 2012-06-29 at 14:48 +0930, Rusty Russell wrote:
> > On Thu, 28 Jun 2012 08:42:56 +1000, Andrew Bartlett <abartlet at samba.org> wrote:
> > > The other concern I have is that libreplace has in general been about
> > > ensuring that the Linux interface is available everywhere, but this
> > > isn't the Linux interface, it is a custom Samba interface being called
> > > by the Linux name.  
> > > 
> > > The distinction matters, because on Linux, we had actually got to the
> > > point where with waf, libreplace actually became a no-op (libbsd is used
> > > for the strl* functions).  (Jelmer was keen on this). 
> > 
> > Nice: it's a commonly-used library so it's not an onerous requirement,
> > and means we don't have to maintain any code (I wish they'd do err, though).
> 
> So far, we still maintain the code (strl*()), we just don't those parts
> of libreplace on Linux (returning us to the empty lib). 
> 
> Certainly we could decide to strictly depend on it, but that hasn't been
> our pattern in the past (and where we have, we have tended to bundle). 

This is kind of the worst of all worlds :( We maintain the code *and* we
use an external library.

Since Jeremy's point is that set*id are flawed, the replacements belong
in libreplace.  He could create another library, but it'd just be
libreplace by another name.  Everything else I can think of seems
awkward: smb_set*id and a big caution not to use raw set*id?

I deeply sympathize with the dream of an NOOP libreplace.  Yet it seems
that it's always going to be needed :(

Cheers,
Rusty.