Re: [PATCH,HURD] Fix grub-probe with userland partition support

[Samuel Thibault <samuel.thibault@ens-lyon.org>]

Vladimir 'φ-coder/phcoder' Serbinenko, le Tue 24 Apr 2012 11:12:11 +0200, a écrit :
> On 24.04.2012 11:00, Samuel Thibault wrote:
> > Vladimir 'φ-coder/phcoder' Serbinenko, le Tue 24 Apr 2012 10:55:21 +0200, a écrit :
> >> On 24.04.2012 01:34, Samuel Thibault wrote:
> >>> +	  if (errno == ENOENT && part >= 5)
> >>> +	    {
> >> This introduces an assumption that if part >=5 is missing then there are
> >> no additional partitions. While this is true for msdos, it's false for
> >> GPT, where such config is common after deleting a partition. Moreover,
> >> APM (apple) partitions often start at 8.
> > Ok, but what do you suggest?  By the very nature of parted-based
> > translators, there is no way to know how many partitions there are.
> Can we know whether given device is partition? If we know that it is we
> can iterate as long as we can, adding some restriction like your 1000
> and issuing error if it's hit since it should never happen.

I don't see how knowing whether the device is partition will help to
know where to stop. A user could very well have all partition nodes in
/dev up to 100, they'll all be partitions. Whether they actually exist
is another story, but as you say, you can't assume anything with the
absence of a partition.

Samuel