=== amitk_ is now known as amitk | ||
AceLan | cking: what kind of regression test should I do when submit SRU patch? | 09:06 |
---|---|---|
cking | AceLan, for the "regression potential" part of a SRU, I just want to see a description of where you think regressions could possibly occur. For fixes such as the PCI power savings, it would be useful to check this on a range of H/W just to see if it works on other devices apart from the ones you are targetting for the fix | 09:07 |
AceLan | cking: got it | 09:10 |
AceLan | cking: about his one "Fix can't disable USB port issue", it's a quirk for specific id, do I need to test it on other machine? | 09:11 |
smb | AceLan, the more a patch changes in code that potentially is run on a lot of HW, the more we get worried. Sorry for rejecting that one request for the second time, but it just feels like a lot of potential to go wrong | 09:11 |
cking | AceLan, I'm very anal about checking stuff, so some checking on other H/W when it comes to some types of fixes is always useful, even the simplest of fixes can sometimes catch one out | 09:12 |
cking | it's called the "can I sleep OK overnight with this fix" sanity check ;-) | 09:13 |
AceLan | smb: thanks for the comment, I'll try to gather more machines for testing | 09:19 |
smb | AceLan, yw, and again sorry for being a pain. But we try to avoid pain on our side ;) | 09:20 |
cking | well, pain on the user's side too if it goes bad | 09:30 |
xnox | in the ADT tests I have | 09:34 |
xnox | 12:18:01 ERROR| [stderr] FAIL: test_061_guard_page (__main__.KernelSecurityTest) | 09:34 |
xnox | 12:18:01 ERROR| [stderr] Userspace stack guard page exists (CVE-2010-2240) | 09:34 |
xnox | failing for kernel, on artful. | 09:34 |
xnox | is this test downloaded by autopkgtest from not the source package? | 09:34 |
xnox | because a package should be able to pass it's autopkgtest. And I don't see how that test is relevant for src:linux triggered by systemd. | 09:34 |
xnox | does it mean artful started to be vulnerable to CVE-2010-2240 ? | 09:35 |
smb | xnox, probably rather the test being confused by the new changes due to CVE-2017-1000364 and needing updates. I believe artful picked up the upstream commits which had some compat issues / regressions in corner cases | 09:43 |
xnox | smb, horum. I am no kernel person and the ADT failure looks scary to the uninitiated. Can the ADT test please be "fixed" to e.g. expectfail / skip that test for now, if it is not in fact regressing? | 09:44 |
xnox | but it seems that update to systemd is not causing this kernel regression. | 09:45 |
smb | xnox, no don't think so to the latter. and the former was was afaik already redirected at the sec team | 09:46 |
Ivanovik | Hi, anyone know the right channel I need to head to if I need NIC internals information? | 11:11 |
Ivanovik | I'm more interested in the datasheet/electronic aspect of it. | 11:11 |
Ivanovik | My goal is to write a driver. | 11:11 |
Ivanovik | I am fairly familiar with the os/kernel interfacing part. | 11:11 |
xnox | smb, hm.... does this mean kernel security update was released without running adt test? | 13:34 |
xnox | (one can run adt test with kvm and cloud image using embargoed / locally built packages) | 13:34 |
smb | xnox, you do realize that the adt test is failing for artful which has a different patch than the rest? | 13:39 |
xnox | smb, yes I do. | 13:41 |
xnox | smb, and my understanding was that linux kernel is built in ppa, with adt tests run, before it is copied into artful-proposed and starts blocking migrations of all packages. | 13:41 |
xnox | e.g. we are failing to migrate gcc, systemd, because of the failing adt test. | 13:41 |
xnox | smb, or e.g. copy linux from ppa to silo, and then to the archive. | 13:43 |
apw | xnox, this was an embargoed cve all things are different | 13:43 |
smb | xnox, this is a devel kernel which is built into proposed like any other package | 13:43 |
xnox | to have a full run of adt tests report, which is automated for the silo ppas. | 13:43 |
apw | xnox, and regardless the one which is affecting your adt tests is not the things we released via security | 13:43 |
xnox | smb, it was built in the PPA for Canonical Kernel team and then copied into artful-proposed. | 13:43 |
apw | xnox, why are we so caring about this one failure, we have adt failures which block things all the time | 13:44 |
xnox | apw, ok. | 13:44 |
apw | xnox, we have a system to cope with it | 13:44 |
xnox | sure. but things that do clog up -proposed are either resolved quickly or removed from proposed. | 13:45 |
xnox | do we need a broken kernel in artful-proposed, if it is not passing adt tests? | 13:45 |
apw | what is it blocking whihc is such a huge pressure ? | 13:45 |
xnox | it's not like it will magically pass, due to changes needed in either kernel or the test-suite? thus that upload is toast, no? | 13:45 |
apw | if you say gcc i will punch you on the nose | 13:45 |
xnox | it's blocking automated migration. | 13:45 |
xnox | systemd | 13:45 |
xnox | gcc for doko; systemd for me. As usual. | 13:46 |
apw | ok so we can hint systemd if that is the only failure | 13:46 |
apw | that is why we have hints | 13:46 |
xnox | apw, but hinting requires humans. | 13:46 |
xnox | i'd rather just remove the src:linux from artful-proposed. | 13:46 |
xnox | rather than badtesting src:linux. | 13:46 |
apw | and you will expect me to not do the same for systemd if it fails for an instant right ? | 13:46 |
smb | and that does not require a human? | 13:47 |
xnox | the test is good; and that src:linux is genuenly toast, no? | 13:47 |
apw | smb, the same humans indeed | 13:47 |
xnox | apw, i expect an upload to fix the systemd in proposed or remove the offending systemd in proposed. | 13:48 |
apw | xnox, indeed, and we are working on that right now | 13:48 |
xnox | obviously we cannot remove src:linux or src:systemd from artful-release. | 13:48 |
xnox | hm it is very odd. | 13:49 |
apw | what is very odd | 13:49 |
xnox | i see test results for 6.11 and 7.12, but not 8.13 and release pocket has 4.10.0-22.24 | 13:49 |
xnox | e.g. systemd migration should be tested against 4.10, not 4.11 | 13:49 |
apw | xnox, yes, if it is not that is a failure in adt | 13:50 |
apw | or some muppet ran all-proposed run | 13:50 |
apw | we should remove that option | 13:51 |
xnox | kernel adt tests are odd; they mostly test the new kernel; and very little test of changed userspace. | 13:51 |
xnox | apw, but do you run adt tests of linux kernel against the new kernel after it is built in the kernel PPA before copying into the devel-proposed? | 13:52 |
apw | xnox, normally yes, unsure if that was done for this one as this was a fix for an emergency CVE | 13:53 |
xnox | because 5.10, 6.11, 7.12, 8.13 of the 4.11 are all failing | 13:53 |
apw | xnox, but as the only people who should be afected are us | 13:53 |
xnox | true. | 13:53 |
apw | xnox, as they are only in -proposed and your testing is against -release | 13:53 |
apw | xnox, so perhaps we need to work out how you got tests against the -proposed version at all | 13:53 |
xnox | yes. | 13:54 |
xnox | meanwhile i'm doing self service request to rerun systemd->linux against release linux, rather than proposed linux. | 13:55 |
=== JanC_ is now known as JanC | ||
=== Ivanovik is now known as John___ |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!