Disks plugin is capped at 2TB size due to field definitions as INT. Changing it to BigInt solves the problem.

On regular builds and thus on Launchpad disks.disks test does not seem to be run anymore and thus I was able to remove the temporary patch via https://salsa.debian.org/mariadb-team/mariadb-10.5/-/commit/bde2cf481fa48a0dd85b9ad40e27ad5005ad1122

See example https://launchpadlibrarian.net/501318051/buildlog_ubuntu-groovy-amd64.mariadb-10.5_1%3A10.5.5-4~ubuntu20.10.1~1602234093.84276d72c.master.next_BUILDING.txt.gz

Tested with an upload to Launchpad where the extended test suite is enabled. Now the test disks.disks ran and failed on all archs:

disks.disks                              w2 [ fail ]

        Test ended at 2020-10-11 23:42:10

CURRENT_TEST: disks.disks

--- /<<PKGBUILDDIR>>/plugin/disks/mysql-test/disks/disks.result	2020-08-07 12:57:37.000000000 +0000

+++ /<<PKGBUILDDIR>>/plugin/disks/mysql-test/disks/disks.reject	2020-10-11 23:42:10.002711946 +0000

@@ -9,4 +9,4 @@

 ) ENGINE=MEMORY DEFAULT CHARSET=utf8

 select sum(Total) > sum(Available), sum(Total)>sum(Used) from information_schema.disks;

 sum(Total) > sum(Available)	sum(Total)>sum(Used)

-1	1

+NULL	NULL

mysqltest: Result length mismatch

https://launchpad.net/~mysql-ubuntu/+archive/ubuntu/mariadb-10.5/+builds?build_text=&build_state=all
https://launchpadlibrarian.net/501653867/buildlog_ubuntu-groovy-amd64.mariadb-10.5_1%3A10.5.5-4~ubuntu20.10.1~1602458170.84f9c0e86.master.next_BUILDING.txt.gz

Upped priority so that this doesn't get past the next 10.x stable release.

was fixed in:

$ git tag --contains 8ec21afc8ed2df3b02815a45624b3287d5ffceae

mariadb-10.5.10

mariadb-10.5.11

mariadb-10.5.12

mariadb-10.5.13

mariadb-10.5.14

mariadb-10.5.4

mariadb-10.5.5

mariadb-10.5.6

mariadb-10.5.7

mariadb-10.5.8

mariadb-10.5.9

mariadb-10.6.0

mariadb-10.6.1

mariadb-10.6.2

mariadb-10.6.3

mariadb-10.6.4

mariadb-10.6.5

mariadb-10.6.6

mariadb-10.7.1

mariadb-10.7.2

mariadb-10.8.1

Where you going to back port it?

I didn't remember I fixed it in 10.5 more than a year ago.

If it keeps failing in buildbot.mariadb.org, yes we should back-port it. There's no reason to have a bug present in earlier versions. I do not remember of my reasoning from back then for not doing it. I'll keep the task open and do it soon. Critical should keep it at the top of the list.

ae3a7d5e4372c4be933c867348d9bc6fca82db24 was the commit in earlier versions