Bug#986581: debian-security-support: logic behind version-based filters

2021-04-26 Thread Holger Levsen
On Mon, Apr 26, 2021 at 07:16:31PM +0200, Sylvain Beucler wrote:
> I think we are all OK with this particular change. Can you review the MR?

yes, I will, ASAP, pinging the the issue every 4 days is too much and will
not speed up things.


-- 
cheers,
Holger

 ⢀⣴⠾⠻⢶⣦⠀
 ⣾⠁⢠⠒⠀⣿⡁  holger@(debian|reproducible-builds|layer-acht).org
 ⢿⡄⠘⠷⠚⠋⠀  OpenPGP: B8BF54137B09D35CF026FE9D 091AB856069AAA1C
 ⠈⠳⣄


signature.asc
Description: PGP signature


Bug#986581: debian-security-support: logic behind version-based filters

2021-04-26 Thread Sylvain Beucler

Hi,

On 16/04/2021 10:41, Sylvain Beucler wrote:

I dropped the version-based check and adapted the test suite:
https://salsa.debian.org/debian/debian-security-support/-/merge_requests/9
pending review with secteam.


I think we are all OK with this particular change. Can you review the MR?

Cheers!
Sylvain



Bug#986581: debian-security-support: logic behind version-based filters

2021-04-16 Thread Sylvain Beucler

Hi Christoph,

Thanks a lot for your precisions,

On 13/04/2021 10:02, Christoph Biedl wrote:

Sylvain Beucler wrote...

We could not find a valid use case for this feature, while it is causing
some missing reports as with 'nodejs', as explained in the above BTS entry.

Did we miss something?


Well, I cannot remember the idea behind the logic, so feel free to do
what you (as a group) consider appropriate.

From guessing, I'd say: The case of "Installed package has a version
number higher than the last supported one" - then the security status of
that package is mostly undefined. Possibly I had backports in mind here:
So if a backport is installed *and* that one has proper security
support, I consider it correct to stay silent about that situation; in
fact, alerting is even wrong because it creates unnecessary noise - that
will educate users to ignore such alerts. But it's indeed hard to tell
whether this situation applies (checking apt-cache policy, checking the
security tracker [please don't], yada yade).


Moreover, I see that:
- backports have no official security support so are never supported
  https://backports.debian.org/FAQ/
- the code incorrectly skips packages with no ("0") version

I dropped the version-based check and adapted the test suite:
https://salsa.debian.org/debian/debian-security-support/-/merge_requests/9
pending review with secteam.


The other feature, being able to end support in advance - well I'd call
that a nice hack and I'd certainly keep it. Although I'd agree it will
rarely be used.


Yes, date-based should be enough for this case.

Cheers!
Sylvain



Bug#986581: debian-security-support: logic behind version-based filters

2021-04-13 Thread Christoph Biedl
Sylvain Beucler wrote...

> I'm investigating an issue in 'debian-security-support' related to how it
> includes/excludes packages by comparing the installed version and the
> supported version, see:
> https://bugs.debian.org/986581

Oy, you brought back a lot of old memories. Not necessarily good ones
but that's not your fault.

Also, I haven't checked the old mails I'd exchanged with the security
team when designing that code. But quite frankly, if a feature is
neither obvious nor documented, it's possibly not worth it.

> We could not find a valid use case for this feature, while it is causing
> some missing reports as with 'nodejs', as explained in the above BTS entry.
> 
> Did we miss something?

Well, I cannot remember the idea behind the logic, so feel free to do
what you (as a group) consider appropriate.

From guessing, I'd say: The case of "Installed package has a version
number higher than the last supported one" - then the security status of
that package is mostly undefined. Possibly I had backports in mind here:
So if a backport is installed *and* that one has proper security
support, I consider it correct to stay silent about that situation; in
fact, alerting is even wrong because it creates unnecessary noise - that
will educate users to ignore such alerts. But it's indeed hard to tell
whether this situation applies (checking apt-cache policy, checking the
security tracker [please don't], yada yade).

The other feature, being able to end support in advance - well I'd call
that a nice hack and I'd certainly keep it. Although I'd agree it will
rarely be used.

Christoph


signature.asc
Description: PGP signature


Bug#986581: debian-security-support: logic behind version-based filters

2021-04-08 Thread Sylvain Beucler

Hello Christoph,

I'm investigating an issue in 'debian-security-support' related to how 
it includes/excludes packages by comparing the installed version and the 
supported version, see:

https://bugs.debian.org/986581

At this point I'm inclined to drop all the version-based logic, because 
when a package is added in security-support-ended.debX, that means there 
is immediate concern about this package, and we already can distinguish 
upcoming and effective end-of-support through dates.


We could not find a valid use case for this feature, while it is causing 
some missing reports as with 'nodejs', as explained in the above BTS entry.


Did we miss something?

Cheers!
Sylvain