svn commit: r274095 - head/release

[Jung-uk Kim]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 11/04/2014 15:24, Dmitry Morozovsky wrote:
> On Tue, 4 Nov 2014, Jung-uk Kim wrote:
> 
>>>> Log: Fix VOLUME_LABEL when BRANCH contains non-alpha
>>>> characters, in particular '-' and '.'.
>>> 
>>> [snip]
>>> 
>>> By the way, what are original reasons to not allow '-' and '_'
>>> in volume names?
>>> 
>>> It looks a bit inconsistent with other file systems volume 
>>> labels...
>> 
>> AFAIK, it is an ISO9660 limitation, i.e., it must be
>> "d-characters".
>> 
>> http://wiki.osdev.org/ISO_9660#String_format 
>> http://wiki.osdev.org/ISO_9660#The_Primary_Volume_Descriptor
>> 
>> In fact, makefs(8) enforces this rule.
> 
> Ah, so iso9660 has stricter rules than, say, xFAT/NTFS and even
> UFS...
> 
> Maybe then we could distinguish file system type given to makefs(8)
> and apply different check/sed rules depending on it?
> 
> I mean: it looks a bit unexpected (if not uncomfortable) to have
> label set up by outer OS accepted by geom_label (and thus created
> in /dev tree) but unable to duplicate it via makefs(8) and similar
> mechanisms...

Our mount_cd9660(8) allows it because there are broken CDs in real world.

http://wiki.osdev.org/ISO_9660#String_format

"Note that not all CDs strictly adhere to the character sets specified
in ISO 9660."

Actually, popular tools, e.g., mkisofs, do not do strict format checking.

Jung-uk Kim
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iQEcBAEBAgAGBQJUWTpjAAoJEHyflib82/FGd3sH+gKDR+7ZyHwZoX2rnfbiweKo
8uYYe50RyNfGvbwztuf6GVl1pP8U3JT3sctTvlDi5teoy2O1LNnFdQucoP0eYMDk
QrybHBoMhsRL8QfnW2VT6NkwjRbojJ39XdItN4Nm4uZXPSG5oI8DjXygwYeEhDXL
Vt+pWf7pHNChcQRjaufoMLCuyWoAWatcJi0tt5DXkwp986E1+B7T9MQLYVzNsTYO
m7b5Byxz9Wxqbax3+pxI8zp6pnaWRgICBvyaDGGCKlsYK5YzRNz2iria4O79oUQT
4eclRZBDnSZNmNnefQhTCA9aTwt4et95bn9ZZfkCE9X83PBRTxmZATtXyhKzRU4=
=cvyS
-----END PGP SIGNATURE-----