Security update 10.1.30-0ubuntu0.17.10.1 regresses smoke test, mariadb not started upon install
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
mariadb-10.1 (Ubuntu) |
Fix Released
|
Undecided
|
Otto Kekäläinen | ||
Artful |
Won't Fix
|
Undecided
|
Unassigned |
Bug Description
[Impact]
- In debian/rules, -pie introduced, yet previously the build was fully hardened. (I believe this is cosmetic, as supposedly the build system defaults to pie anyway)
- There used to be a default client socket set, but not anymore.
- Most importantly, the dh_systemd_start workarounds were dropped; yet are still needed on artful, as otherwise after server installation the unit is not started / not running
- Due to last one, autopkgtests are failing - see http://
[Solution]
- perform partial revert of packaging changes, back to what it was before (all the hacks & workarounds included)
[Testcase]
- autopkgtests should pass for mariadb-10.1 package, specifically the smoke test case
[Regression Potential]
Reverting packaging changes introduced in the last update; back to what they were before and what they are in the release pocket. This revert may have been done incorrectly, thus yielding more maintainer script bugs, but interactive testing of this revert appears to be functioning correctly.
[Original Bug report]
10.1.30-
Setting up mariadb-
Setting up mariadb-
Setting up mariadb-client-10.1 (10.1.30-
Setting up mariadb-server-10.1 (10.1.30-
Created symlink /etc/systemd/
Created symlink /etc/systemd/
Created symlink /etc/systemd/
Setting up autopkgtest-satdep (0) ...
Processing triggers for libc-bin (2.26-0ubuntu2.1) ...
Processing triggers for systemd (234-2ubuntu12.3) ...
(Reading database ... 69452 files and directories currently installed.)
Removing autopkgtest-satdep (0) ...
autopkgtest [00:52:38]: test smoke: [------
+ mysql
ERROR 2002 (HY000): Can't connect to local MySQL server through socket '/var/run/
autopkgtest [00:52:39]: test smoke: -------
autopkgtest [00:52:39]: test smoke: - - - - - - - - - - results - - - - - - - - - -
smoke FAIL non-zero exit status 1
Looks like the server is not started ?
It appears so, at least reproducible in lxd container. There are maintainer scripts to start init.d script, however that is not done for systemd. I do not se for example deb-systemd-helper start mariadb.service at all. Possibly regression due to using a new/different debhelper during package build.
Changed in mariadb-10.1 (Ubuntu): | |
importance: | Undecided → Critical |
summary: |
- Security update 10.1.30-0ubuntu0.17.10.1 regresses smoke test + Security update 10.1.30-0ubuntu0.17.10.1 regresses smoke test, mariadb + not started upon install |
information type: | Public → Public Security |
tags: | added: id-5ab3c4d27fa5364dd97ee6d2 |
Changed in mariadb-10.1 (Ubuntu): | |
importance: | Critical → Undecided |
assignee: | nobody → Otto Kekäläinen (otto) |
Changed in mariadb-10.1 (Ubuntu Artful): | |
status: | Confirmed → Won't Fix |
It is easiest if we list the commits that caused the regressions, and then selectively revert them.
Here are some I found quickly: /salsa. debian. org/mariadb- team/mariadb- 10.1/commit/ b4fe91faee29bc1 597e2876e1c2cc6 bdd162fb66 /salsa. debian. org/mariadb- team/mariadb- 10.1/commit/ 9f49e4b494f3dad 8c403972996f7a1 ebceb4b34f
- -pie https:/
- dh_systemd_start removal https:/
At least the commits above are well documented and will help to understand if reverting them is the truly correct thing to do.