Bug#776021: unblock: cryptsetup/2:1.6.6-5
Hi,
Am 29.01.2015 um 19:26 schrieb Cyril Brulebois:
> Niels Thykier <niels@thykier.net> (2015-01-29):
>> Ack from RT, adding KiBi to CC for a d-i ack.
>>
>> Quoting the rest of the mail in full for KiBi's convenience.
>
> (ta)
>
>>> I just uploaded cryptsetup 2:1.6.6-5 to unstable, fixing one single
>>> reported bug, but also another - unreported - much worse bug. The report is:
>>>
>>> Bug#773456: [cryptsetup] invoke-rc.d not found on boot
>>>
>>> The severity of this bug is *just* 'normal'. But while fixing the bug, I
>>> discovered a far worse bug in the same line of code, which I'd really
>>> like to have fixed in time for jessie. I'm going to explain:
>>>
>>> */ The recently (in last upload, 2:1.6.6-4) introduced grep regex on
>>> outputof 'status cryptdisks-udev ...' is insufficient. On systems
>>> that use an init system != upstart but have upstart-bin installed,
>>> this grep regex can lead to infinitive loops in the boot process.
>>> [If you want me to, I can open a separate bugreport with rc severity
>>> regarding this problem.]
>>> */ Apart from that, the recently (in last upload, 2:1.6.6-4) introduced
>>> check for an upstart environment is broken. This will result in ugly
>>> error messages on every system with encrypted devices and separate
>>> /usr partition. (#773456)
>>> */ The fix introduced in this upload is very unintrusive. It only
>>> changes the (before broken) logic to detect an upstart environment
>>> and additionally fixes the (before insufficient) grep regex to
>>> detect a currently running cryptdisks-udev (upstart) init script.
>>>
>>> The debdiff is attached and here's the full changelog for this upload:
>>>
>>> cryptsetup (2:1.6.6-5) unstable; urgency=high
>>>
>>> * debian/cryptdisks.functions: fix the precheck for ubuntu+upstart
>>> before invoking 'status cryptdisks-udev'. (closes: #773456)
>>> * debian/cryptdisks.functions: fix the insufficient grep regex for
>>> detecting a running cryptdisks-udev (upstart) init script.
>>>
>>> -- Jonas Meurer <mejo@debian.org> Thu, 22 Jan 2015 21:22:08 +0100
>>>
>>> unblock cryptsetup/2:1.6.6-5
>>>
>>> Cheers,
>>> jonas
>>>
>>> -- System Information:
>>> [...]
>
> This was quite unclear from the BTS when I looked earlier…
> | kibi@arya:~/debian-installer/scripts/testing-summary$ svn log -c 69584
> | ------------------------------------------------------------------------
> | r69584 | kibi | 2015-01-29 09:04:04 +0100 (Thu, 29 Jan 2015) | 1 line
> |
> | Add cryptsetup as unwanted.
> | ------------------------------------------------------------------------
> | kibi@arya:~/debian-installer/scripts/testing-summary$ svn diff -c 69584
> | Index: unwanted-packages
> | ===================================================================
> | --- unwanted-packages (revision 69583)
> | +++ unwanted-packages (revision 69584)
> | @@ -8,3 +8,4 @@
> | iprutils 2.4.5-1 New upstream release, no ACK from the maintainer.
> | libevdev 1.3+dfsg-2 RT-declined unblock request (#771014).
> | wide-dhcpv6 20080615-13 Too many changes.
> | +cryptsetup 2:1.6.6-5 Ubuntu-specific changes for a normal bug.
>
> Why isn't the BTS updated with the information contained in the unblock
> request? It'd be very helpful to avoid having incorrect or insufficient
> information in the BTS, if at all to avoid wasted efforts…
I'm sorry for that. As I tried to explain in the unblock request,
bugreport #773456 was not exactly about the critical problem I
introduced with 2:1.6.6-4. But I should have opened an RC bug myself
before doing the upload to make this more clear to you. I apologize for
not doing that. Now it's clear to me that this would be the correct way
to go, especially in that phase of release freeze - and that I would
have saved some hassle on your side that way.
> Since we have this:
> cryptsetup-udeb: /lib/cryptsetup/cryptdisks.functions
>
> I'm going to ask whether anyone has tested this new cryptsetup in a d-i
> context to make sure the fix isn't triggering more fun elsewhere…
>
> Failing at least basic testing, I'm not exactly excited with a quick
> migration to testing.
It's a one-line fix for the init scripts. cryptdisks.functions is
sourced by the init scripts and by crypdisks_st{art,op}, it's not used
anywhere else. I cannot imagine any interference with debian-installer.
Actually the only relevant change is that the precheck for a
ubuntu-specific upstart initscript changed from
> invoke-rc.d --quiet cryptdisks-udev status && ...
to
> [ -x /sbin/status ] && ...
The one-character change following in the grep regex is only relevant to
upstart-based ubuntu systems anyway.
But if you want me to, I could do basic testing. I guess that testing
the daily-built images from sid would be the right thing in that case?
Cheers,
jonas
Reply to: