Comments from IRC after I posted this proposal:
20:20 < Ryan52> bubulle: your mdadm rescue patch does nothing, right? cause the
man page says that the default is
--auto=yes. and the auto arguement is unrelated, aiui.
20:21 < bubulle> Ryan52: ah, crap.
20:21 < bubulle> Well, I ws saying that my patch is naive...:-)...you're just
proving me right
20:21 < Ryan52> :)
20:21 < bubulle> so, the real patch is (probably still naive) --auto=no
20:21 < Ryan52> no.
20:21 < Ryan52> that whole line should probably be removed. and the one before
it if /tmp/mdadm.conf isn't used
elsewhere...
20:22 < bubulle> Ryan52: would be good if you can update the bug, then (and
better if you're in the position of
testing)
20:22 < fjp> That's all crap IMO.
20:23 < Ryan52> fjp, what is?
20:23 < fjp> What needs to happen is addition of sanity checks that the
generated /tmp/mdadm.conf is sane.
20:23 < fjp> And only if it isn't we should prevent mounting.
20:24 < fjp> What you're proposing now is dropping functionality without
providing an alternative.
20:24 < fjp> And any alternative is going to mean added strings...
20:24 < bubulle> yes, this is what I call a "naive" proposal
20:26 < fjp> Although rather nasty, the real problem here is basically that the
system was extremely dirty with
loads of old no longer valid raid IDs lying around.
20:27 < fjp> Given that Martin is the mdadm maintainer, the chance that he'd
run into this is a huge factor
larger than that anybody else will.
20:28 < Ryan52> okay, well, I can't help with that (at least this weekend..)
20:29 < bubulle> fjp: so, actually, your advice would be to ignore that issue
for lenny (which is anyway what we
will have to do), add it to the errata, and defer a clean fix
for squeeze?
20:29 < bubulle> such as proposing the user to auto-assemble the RAID arrays or
not, or whatever solution that
adds user interaction
signature.asc
Description: Digital signature
