Re: Comments on the psmisc patch?

[Guillem Jover <guillem@debian.org>]

On Mon, 2011-10-10 at 23:20:15 +0200, Svante Signell wrote:
> In the patch of fuser.c I rewrote the xreadlink() function, since the
> code in the Alioth patch at
> http://www.gnu.org/software/hurd/hurd/porting/guidelines.html 
> does not work :-( 

I've mentioned this before, for symlinks the ideal solution is to
lstat(2) and then allocate and readlink(2) using st_size + 1.

In this case this does not seem desirable (at least on expandpath)
because the code is trying not to use stat(2) due to NFS issues.

> The last change in fuser.h, not including lists.h jumps over large parts
> of the code, but since the /proc interface is only partially supported
> on Hurd, not much harm is done. An example is: /proc/self/mountinfo

I think the correct way to fix this is to just partially revert to the
previous upstream code, so that the mountinfo ugliness, _LISTS_H_,
expandpath and all its accompanying PATH_MAX issues get conditionalized
for Linux systems only. There's no point in all that code on other
systems really, and more so when upstream has conditionalized and
disabled by default the timeout fork stat (for the NFS issue) on
latest git.

thanks,
guillem