Bug in bsdinstall (fs found where not present)

[Nathan Whitehorn]

On 08/29/13 22:39, Mehmet Erol Sanliturk wrote:
>
>
>
> On Thu, Aug 29, 2013 at 11:09 PM, Nathan Whitehorn
> <nwhitehorn at freebsd.org <mailto:nwhitehorn at freebsd.org>> wrote:
>
>     On 08/29/13 17:02, Warren Block wrote:
>     > On Thu, 29 Aug 2013, Nathan Whitehorn wrote:
>     >
>     >> On 08/29/13 14:04, Warren Block wrote:
>     >>>> From a 9.2-PRERELEASE snapshot, go into the shell, create a
>     GPT disk
>     >>> layout with a bunch of partitions for filesystems and swap.
>      Exit the
>     >>> shell and run the installer.
>     >>>
>     >>> Go through each partition setting a mount point.  Tell
>     bsdinstall to
>     >>> continue.  It reports that the / partition has a preexisting
>     >>> filesystem (it does not, in fact; this disk had a mishmash of
>     MBR and
>     >>> NTFS on it).
>     >>>
>     >>> Tell bsdinstall to continue anyway.  It does, and then reports
>     that it
>     >>> can't mount /dev/ada0p2 on /mnt, presumably because, contrary
>     to the
>     >>> misleading and incorrect error message, there is no filesystem on
>     >>> there.
>     >>>
>     >>> The install fails, try again, entering all the mount points,
>     and it
>     >>> will fail the same.
>     >>>
>     >>> Short term solution: newfs the / partition, so there really is a
>     >>> filesystem there for bsdinstall to detect and warn about.  Then it
>     >>> works.
>     >>
>     >> bsdinstall has no way to detect whether or not you already have
>     UFS in a
>     >> freebsd-ufs file system. It assumes, when not given contrary
>     >> information, that a partition that exists is initialized. There
>     does not
>     >> seem to be a way around this. If you have any ideas, those would of
>     >> course be helpful.
>     >
>     > file(1) works well for detecting filesystems.
>     >
>     > For that matter, what is bsdinstall doing now that makes it say
>     there
>     > is a filesystem on a partition?  Maybe the message is misleading.
>
>     What that actually means is that the partition exists. (file doesn't
>     work on block devices, by the way) I'm happy to change the error
>     message. The default behavior is that, like partitioners on all other
>     operating systems, it treats creating partitions and running newfs as
>     intimately linked activities -- similarly, that the type marked in the
>     partition table is the actual filesystem type. Intermediate cases are
>     very hard to detect reliably.
>     -Nathan
>     _______________________________________________
>
>
>
>
> I am installing many different operating systems .
>
> One of the important problems is when
>
> "Use entire disk"
>
> is selected , some of the installers are still searching valid
> partitions on disk
> and failing miserably because there does not exist any one ( because
> unit is new or corrupted ) or there are some partitions , etc.
> remained from another different operating system .
>
> My opinion is that the first question should be to select an option among
>
> "Use entire disk" or
> "Use existing file systems" .
>
> alternatives , only search valid file system(s) when "Use existing
> file systems" is
> selected .
>
> When "Use entire disk" is selected , directly apply file systems
> creations by just
> after determining the geometry of the unit under consideration .

I agree with you -- that is exactly what the installer does in this case.
-Nathan