[gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[schism]

Normally I don't take to the skies with this sort of thing, but this was
ridiculous.  In what I am sure was a well-intentioned change, cryptsetup
recently silently lost its default USE=static status.  For one of the
many people actually using cryptsetup in an initrd, this came as a
lovely surprise this morning as I found my new kernel/initrd completely
unbootable.

To compound matters, udev-181's incompatibility with my old kernel (no
DEVTMPFS, separate /usr), and my primary system was completely
unbootable.

Please - if you're going to significantly change functionality that
(whether correct or not) people have come to depend on, have the decency
to at least publish a news article on it.  I may find 'eselect news'
annoying, but I at least read it.  Before we get to bikeshedding, I'm
well aware there were multiple ways to have prevented this, but none of
them were in place [1].  What was in place was default behavior that had
been there for a very long time.


[1] For one, genkernel should bomb out if it can't comply with a
command-line arg instead of just putting non-alert text up.

Re: [gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[Maxim Kammerer]

On Thu, Mar 22, 2012 at 15:48,  <schism@subverted.org> wrote:
> Please - if you're going to significantly change functionality that
> (whether correct or not) people have come to depend on, have the decency
> to at least publish a news article on it.

Previous stable cryptsetup had an "if use dynamic" warning in
pkg_setup(), which was removed instead of being converted to "if ! use
static" in the new stable. No idea if that was intentional.

-- 
Maxim Kammerer
Liberté Linux (discussion / support: http://dee.su/liberte-contribute)

Re: [gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[Alexandre Rostovtsev]

On Thu, 2012-03-22 at 07:48 -0600, schism@subverted.org wrote:
> Normally I don't take to the skies with this sort of thing, but this was
> ridiculous.  In what I am sure was a well-intentioned change, cryptsetup
> recently silently lost its default USE=static status.  For one of the
> many people actually using cryptsetup in an initrd, this came as a
> lovely surprise this morning as I found my new kernel/initrd completely
> unbootable.
> 
> To compound matters, udev-181's incompatibility with my old kernel (no
> DEVTMPFS, separate /usr), and my primary system was completely
> unbootable.
> 
> Please - if you're going to significantly change functionality that
> (whether correct or not) people have come to depend on, have the decency
> to at least publish a news article on it.  I may find 'eselect news'
> annoying, but I at least read it.  Before we get to bikeshedding, I'm
> well aware there were multiple ways to have prevented this, but none of
> them were in place [1].  What was in place was default behavior that had
> been there for a very long time.
> 
> 
> [1] For one, genkernel should bomb out if it can't comply with a
> command-line arg instead of just putting non-alert text up.

There is already a bug open about this issue:
https://bugs.gentoo.org/show_bug.cgi?id=409277

However, I must still thank you for alerting the list about it; you
saved lots of peoples' systems (probably including mine, if I had not
read this message) from being rendered unbootable.

-Alexandre Rostovtsev.

Re: [gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[Sebastian Pipping]

On 03/22/2012 03:20 PM, Alexandre Rostovtsev wrote:
>> [1] For one, genkernel should bomb out if it can't comply with a
>> command-line arg instead of just putting non-alert text up.
> 
> There is already a bug open about this issue:
> https://bugs.gentoo.org/show_bug.cgi?id=409277

With that bug fixed by now is there still need for a news entry?

Best,



Sebastian

Re: [gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[Samuli Suominen]

On 03/29/2012 12:58 PM, Sebastian Pipping wrote:
> On 03/22/2012 03:20 PM, Alexandre Rostovtsev wrote:
>>> [1] For one, genkernel should bomb out if it can't comply with a
>>> command-line arg instead of just putting non-alert text up.
>>
>> There is already a bug open about this issue:
>> https://bugs.gentoo.org/show_bug.cgi?id=409277
>
> With that bug fixed by now is there still need for a news entry?
>
> Best,
>
>
>
> Sebastian
>

It doesn't look fixed to me, only temporarily worked around to force USE 
static on cryptsetup.
Should be kept open until genkernel is fixed to generate the initramfs 
in a way it doesn't need the static cryptsetup.

I've been told dracut is able to handle this.   Unverified.

- Samuli

Re: [gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[Amadeusz Żołnowski]

[-- Attachment #1: Type: text/plain, Size: 211 bytes --]

Excerpts from Samuli Suominen's message of 2012-03-29 19:59:17 +0200:
> I've been told dracut is able to handle this.   Unverified.

Dracut doesn't need anything built static.


-- 
Amadeusz Żołnowski

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[Samuli Suominen]

On 03/30/2012 10:56 AM, Amadeusz Żołnowski wrote:
> Excerpts from Samuli Suominen's message of 2012-03-29 19:59:17 +0200:
>> I've been told dracut is able to handle this.   Unverified.
>
> Dracut doesn't need anything built static.

Thanks for verifying. Expected nothing less. I've reopened the bug[1] 
for genkernel so it doesn't get lost...

[1] http://bugs.gentoo.org/409277

Re: [gentoo-dev] Arbitrary breakage: sys-fs/cryptsetup

[schism]

On Thu, Mar 29, 2012 at 11:58:17AM +0200, Sebastian Pipping wrote:
| With that bug fixed by now is there still need for a news entry?

I don't think I make that call, but if the current plastering-over would
generally keep existing setups from breaking, I'm happier.