Stable release 3.4.8 available

[Julien BLACHE]

Michael Tautschnig <mt at debian.org> wrote:

Hi,

> Would you have a disk config available, where you've seen breakage? I'd then
> check it... If not, could you give a slightly more verbose description of the
> configuration such that I could try to come up with a supposedly broken one?

I can provide that tomorrow; basically having an extended partition that
takes all the remaining disk space should trigger the bug.

>>  - it is compatible with previous releases when preserving msdos
>>    disklabels;
>
> The current code should ensure this.

I read the patches after the FAI meeting and at first glance I had my
doubts. As I haven't put it to the test, I can't be affirmative; see
below, though.

>> As proposed already, let's just forget about alignment and msdos
>> disklabel. It's not worth the hassle.
>
> I'd love to be able to say that, but it seem that people insist on using msdos.

msdos is a fscking kludge, getting that to work all the while satisfying
"arbitrary" alignment constraints *and* being backward-compatible with
previous releases doesn't seem possible. The biggest issue here is
extended partitions, but then any partition with a fixed size *will*
come up with a different size with the alignment code.

The previous/current code cannot be emulated by setting a fixed
alignment value; primary and extended partitions have very different
behaviours.

Anybody still using msdos disklabels and requiring aligned partitions
for performance should move to GPT *YESTERYEAR*. If the storage isn't
distinct from the system volume and the system firmware refuses to boot
without an msdos disklabel, then use gptsync and you'll be just fine
(hello HP BIOS from the 80ies).

> Well, maybe all we have to do is make gpt the default? Well, but that shouldn't
> go in stable either :-)

Well, maybe we should just leave stable alone? ;)

>> I'm swamped right now, I'm afraid won't find the time to dig into this
>> for weeks if not months, otherwise I'd have raised the issue earlier
>> with patches :/
>
> Well, you did raise the issue earlier on anyway and did provide patches! It's
> just that I tried to strive for a seemingly more general and beautiful solution.
> But obviously beauty is in the eye of the beholder.

msdos, beauty: pick one ;) It's really nobody's fault if msdos-related
code gets ugly.

Though I recently had to dig into Sizes.pm and as soon as I'll have some
time to dig into that again, I'm going to send a patch to clean up a bad
case of variable reuse that had me go WTF??? :)

Could be a good way to keep me busy on the train this week, although not
sure digging into that code in the wee hours is a good idea %) Hmm.

JB.

-- 
 Julien BLACHE - Debian & GNU/Linux Developer - <jblache at debian.org> 
 
 Public key available on <http://www.jblache.org> - KeyID: F5D6 5169 
 GPG Fingerprint : 935A 79F1 C8B3 3521 FD62 7CC7 CD61 4FD7 F5D6 5169