Re: exec server and /dev/fd/N

[olafBuddenhagen]

Hi,

On Wed, May 26, 2010 at 12:29:06PM +0200, Emilio Pozuelo Monfort wrote:
> On 26/05/10 09:32, olafBuddenhagen@gmx.net wrote:

> > IIRC my major concern with this approach was that this way, the
> > original task has to guess what file name will be valid in the new
> > task's context... I don't remember though what my conclusion was
> > regarding this being better or worse than doing the guessing in exec
> > :-(
> 
> The point is that the original task doesn't need to do any guessing,
> since it knows the file name.

It knows the file name in its *own* namespace; but it might not
necessarily be the same in the *new* task...

But as Fredrik pointed out, this is probably not a serious problem.

> > just copy the new .defs to /include/hurd before building the new
> > glibc, which you can then use while building the new Hurd.
> 
> Sounds good for testing, but probably not a good idea for e.g. the
> Debian packages.

Eh? It's the proper way: build a new glibc package with the new .defs;
and then build a new Hurd package with the new glibc. It's always been
done like that. There is even a special target for installing the
interface definitions in the Hurd Makefile, precisely for this purpose.

-antrik-