Triage and Migration for Friday

Christian Ehrhardt christian.ehrhardt at
Fri Feb 28 09:30:05 UTC 2020

Subject: Triage and Proposed-Migration Report for Friday

## Triage ##

I've got 14 bugs to look at (and did a few more of paride and andreas that
got punted due to feature freeze frenzy).

A few needed some minor triage, but none was needing further big effort.

I picked some more waiting triage days and got 32 more bugs.
A few even got into out queues that need further work, of those worth
mentioning are two:
This seems to be a change due to the merge that should be analyzed before
20.04 is final.
@Rafael I assigned you for now to take a look as you did the merge.
IPv4 issues with nmbd, I feel like I had heard Andreas mention that.
@Andreas - could you answer on that bug if it is a known issue (maybe a
dup) or something new we really need/want to look at?

In general the subscription and server-next queue are rather full now.
Let us maybe next week sit down and assign bugs from this queue to team
members to get this down - especially for bugs that should/could be fixed
in 20.04 before release to then SRU things while the release is in its
latest hardening steps where it is hard to upload 20.04 packages anyway.

## Proposed Migration ##

I've actually worked on this for a few days now unstucking libnbd/nbdkit as
well as helping to get the new set of updates for azure-cli done. Nothing
big to mention individually.

Therefore after talking with bryce I was taking a look at arm build issues
of to help the php
migration to go on.

Seeing a build time test fail on armhf&arm64:

TEST 35/132 [tests/034.phpt]
XFAIL Unserialize invalid random data [tests/034.phpt]   XFAIL REASON:
TEST 36/132 [tests/035.phpt]
FAIL Profiling perf test. [tests/035.phpt]
TEST 37/132 [tests/040.phpt]
SKIP broken random data test [tests/040

On an arm64 cloud instance I uprgaded to focal-proposed and ran the build
And of course for the loacl test there I got:
Number of tests :  132               128
Tests skipped   :    4 (  3.0%) --------
Tests warned    :    0 (  0.0%) (  0.0%)
Tests failed    :    0 (  0.0%) (  0.0%)
Expected fail   :    1 (  0.8%) (  0.8%)
Tests passed    :  127 ( 96.2%) ( 99.2%)

While the one breaking on LP has:
Number of tests :  132               128
Tests skipped   :    4 (  3.0%) --------
Tests warned    :    0 (  0.0%) (  0.0%)
Tests failed    :    1 (  0.8%) (  0.8%)
Expected fail   :    1 (  0.8%) (  0.8%)
Tests passed    :  126 ( 95.5%) ( 98.4%)


Tests ran against 7.4 as intended (and as in the failing case):
PHP         : /usr/bin/php7.4
PHP_SAPI    : cli

As silly as it is, I was retrying the LP build to see if it would change.

FYI and probably related:
- (other) 7.4 FTBFS fails on Debian:
- upstream report old but still ongoing

On the latter they mention (
that the test is dependent on the build-env "speed" and might fail on slow
This might be exactly what differs between the build env on LP and my 8
core arm64 test box.

I can rerun the tests in my environment via
$ cd build-7.4
$ make test

As-is I pass the test in 9/10 runs (but not always) of the cases usually
running in 15-17 seconds.

With background load hogging my CPU/kernel/Randomness (stress-ng -c 8
--x86syscall 8 --urandom 8 --rdrand 8 --metrics-brief --getrandom 8) I get
reliable fails 5/5 runs in ~30 seconds runtime.

At the builders a run takes ~20 seconds and have less CPUs.
So this might really just be unreliable to be run in build environments.
Let us skip the test and be good with it.

@Bryce - please try it with this debdiff (attached)

Christian Ehrhardt
Staff Engineer, Ubuntu Server
Canonical Ltd
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fix-php-msgpack.debdiff
Type: application/octet-stream
Size: 1963 bytes
Desc: not available
URL: <>

More information about the ubuntu-server mailing list