Re: /proc/self/fd support on Hurd

[Svante Signell]

On Tue, 2020-02-11 at 06:16 -0800, Samuel Thibault wrote:
> Florian Weimer, le mar. 11 févr. 2020 15:06:43 +0100, a ecrit:
> > > Samuel Thibault, le mar. 11 févr. 2020 05:47:30 -0800, a ecrit:
> > > > Florian Weimer, le mar. 11 févr. 2020 14:39:23 +0100, a ecrit:
> > > > > Does Hurd support /proc/self/fd, in the sense that the POSIX functions
> > > > > in glibc (such as openat) can deal with such paths?
> > > > 
> > > > Not yet but that could be added quite easily with some magic
> > > > translation.
> > > 
> > > Ah, actually it's already supported in /dev/fd (but we can add a symlink
> > > in /proc/self for compatibility).
> > 
> > Does this mean if we generate paths starting with /dev/fd/, we already
> > have this functionality today?

There are already symlinks there:

ls -l /proc/self/fd/
total 0
lrwx------ 1 srs srs 64 Feb 11 14:48 0 -> /dev/pts/1
lrwx------ 1 srs srs 64 Feb 11 14:48 1 -> /dev/pts/1
lrwx------ 1 srs srs 64 Feb 11 14:48 2 -> /dev/pts/1
lr-x------ 1 srs srs 64 Feb 11 14:48 3 -> /proc/2872/fd

ls -l /dev/fd/
total 0
lrwx------ 1 srs srs 64 Feb 11 17:38 0 -> /dev/pts/1
lrwx------ 1 srs srs 64 Feb 11 17:38 1 -> /dev/pts/1
lrwx------ 1 srs srs 64 Feb 11 17:38 2 -> /dev/pts/1
lr-x------ 1 srs srs 64 Feb 11 17:38 3 -> /proc/2982/fd

ls -l /dev/pts/
total 0
crw--w---- 1 srs  tty  136, 0 Feb 11 14:38 0
crw--w---- 1 srs  tty  136, 1 Feb 11 17:39 1
c--------- 1 root root   5, 2 Dec 13 11:55 ptmx