[MDEV-23090] Sid builds of ha_mroonga lost liblz4.so dependency Created: 2020-06-29 Updated: 2023-04-27 |
|
| Status: | Stalled |
| Project: | MariaDB Server |
| Component/s: | Platform Debian, Storage Engine - Mroonga |
| Affects Version/s: | 10.3, 10.4, 10.5 |
| Fix Version/s: | 10.4, 10.5 |
| Type: | Bug | Priority: | Major |
| Reporter: | Elena Stepanova | Assignee: | Otto Kekäläinen |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | None | ||
| Description |
|
Since the last releases, ha_mroonga.so has lost one of dependencies on Sid, both x86 and amd64:
As I understand serg, it may be a bug, a loss of the library. If so, it was probably caused by the recent rebuild of VMs (TODO-2441). |
| Comments |
| Comment by Daniel Bartholomew [ 2020-06-30 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
The rebuilt sid build VMs have liblz4-dev installed. It was actually automatically installed when I ran 'sudo apt build-dep mariadb-server' when creating the build VMs. So I'm not sure why it disappeared from the dep list for ha_mroonga.so. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2020-07-03 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Indeed, the library is there, and moreover, other components still link with it, it's only mroonga that doesn't. So I suppose the difference was still caused by the VM upgrade, but not by the lack of library. Maybe otto will be interested to investigate. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Otto Kekäläinen [ 2020-07-04 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Thanks for reporting elenst! Is the snippet above from a command you ran manually or do we have some linked libraries listing done automatically in buildbot? I was wondering how to test this automatically and/or search from old logs. It is good that we catch these failures during development building against latest upstream dependencies, otherwise it is much more difficult to fix if they are detected only after official releases. Eventually what is in Debian unstable is what is in Debian stable and Ubuntu stable in 1-2 years. The latest upload of a new liblz4-dev to Debian unstable was in November 2019: https://tracker.debian.org/pkg/lz4 Would this match with the builder image update frequency? The liblz4 was not an explicit dependency until it was added in April 2020 in 5cdf245d7e2ab339ad3dba0dbbb591ab80e0dad0. The git history of storage/mroonga does not have any commits related to this to what I can see. A more likely culprint is in the lz4 detection code recently changed by serg:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Otto Kekäläinen [ 2020-07-04 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Related:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2020-07-04 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
We do now, recently added upon Sergei's request.
The mismatches are shown as a failure of minor-upgrade-all test, where at the end of the test you will see the output similar to the snippet. Here is an example: However, this particular problem won't be visible anymore in the new builds, because I have updated the baseline to match the new status.
I don't see how it is possible. This change hasn't been released yet, and as I said in the earlier comment, even when mroonga is built on the new Sid images from the already released 10.3.23, it doesn't link anymore with liblz4, even though it did when it was built from the same sources before the release. At least it didn't when I tried to build it on the current Sid image, using the same steps as buildbot does. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Otto Kekäläinen [ 2020-07-04 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Thanks for the pointers! As reference, status in Debian.org repositories:
History for 10.5.x releases on mariadb.org for Debian Sid:
Indeed, regression in the 10.5.4 sid release. No regression on the Buster and Stretch releases:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Otto Kekäläinen [ 2020-07-05 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
I did not find what the regression is and I cannot put more time on this right now as there are so many open PRs I need to champion/refresh/refine to fix urgent bugs. I'll return to this later, or maybe faust can try hunting what the regression in Sid was? | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Otto Kekäläinen [ 2020-07-05 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
This is maybe related: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=887655 Other notes:
I am confused on that conclusions to draw here. Need to dig up more in the CMake files on how Mroonga builds and how it includes lz4... |