Tests that fail on 10.2

[midenok]

These tests should be excluded from run because they fail on 10.2 branch:

tokudb.rows-32m-seq-insert
rows-32m-rand-insert

[midenok]

innodb_onlinealter_encryption can sometimes fail (probably bug not related to versioning).

--- /home/midenok/src/mariadb/hagrid/src/mysql-test/suite/encryption/r/innodb_onlinealter_encryption.result     2016-08-31 05:03:13.595401621 +0000
+++ /home/midenok/src/mariadb/hagrid/src/mysql-test/suite/encryption/r/innodb_onlinealter_encryption.reject     2016-09-13 00:03:24.096927888 +0000
@@ -31,7 +31,7 @@
 # t1 yes on expecting NOT FOUND
 NOT FOUND /foobar/ in t1.ibd
 # t2 ... on expecting NOT FOUND
-NOT FOUND /temp/ in t2.ibd
+FOUND /temp/ in t2.ibd
 # t3 ... on expecting NOT FOUND
 NOT FOUND /barfoo/ in t3.ibd
 # t4 ... on expecting NOT FOUND

[midenok]

Eugene, can you please look if these 2 tests rows-32m-... are failing for you?

[kevgs]

Why do we care of tokudb?

[midenok]

From skype history:

[14-Sep-16 22:19:24] Alexander Krizhanovsky: Alexey, did you analyze the reasons for #13 failures?
[14-Sep-16 23:13:39] Aleksey Midenkov: "Alexey, did you analyze the reasons for #13 failures?" No. But we don't need to analyze bugs not related to versioning, do we? Daniel, how tokudb.rows-32m-seq-insert and tokudb.rows-32m-rand-insert run on your host?
[14-Sep-16 23:17:05] Alexander Krizhanovsky: How can we be sure that the issues aren't related to versioning if we don't analyze them?
[14-Sep-16 23:17:24] Aleksey Midenkov: Because they fail on 10.2 branch.
[14-Sep-16 23:17:37] Alexander Krizhanovsky: If they really aren't related to versioning, then it would be good to collaborate w/ the guys and report the bugs
[14-Sep-16 23:17:53 | Edited 23:18:27] Daniel Fiala: I disabled tokudb compilation since it is causing some headaches. I think tokudb is broken with source we forked.
[14-Sep-16 23:19:00] Aleksey Midenkov: Fair enough. That's one of the reasons why I mentioned them in issue.
[14-Sep-16 23:20:03] Aleksey Midenkov: Daniel, could you please run them on 10.2 branch so we could be sure that it's not my special environment that fails them.
[15-Sep-16 00:18:07 | Edited 00:19:54] Daniel Fiala: I tried both tests and I had to stop them before they finished. Process mysqld did consume a lot of memory and tests ran for a really long time. I tried our fork with branch 10.2 .

Basically, we do care about all tests, because any of them can trigger failure in our code. These 2 particular tests fail not because of our code, but nevertheless it's worth to collect as much info as possible to decide what to do next: whether disable them, report to upstream or both.

I've already disabled them in trunk, so you may checkout vanilla 10.2 (from your repo) to run them.

[midenok]

Also, be comfortable to run any tests at any time on any branches. Personally, I have following configuration:

alias mtrb='mtr --big-test --big-test'
alias mtrf='mtr --big-test'
alias mtrh='mysql-test-run --help | less'
alias mtrm='mtr --suite=main'
alias mtrx='mtr --extern socket=/home/midenok/src/mariadb/hagrid/build/mysql-test/var/tmp/mysqld.1.sock'
mtr ()
{
    ( mkdir -p "$log_dir";
    cd "$log_dir";
    exec mysql-test-run --force --max-test-fail=0 --suite-timeout=1440 "$@" 2>&1 | tee -a mtr.log )
}

[avodaniel]

Let me at other part of Skype history:

On 15/09/16, at 09:09, Daniel Fiala wrote:

Hi. I tried to run both tests for unlimited amount of time and they both failed. So it seems that there is really something odd in tokudb implementation.

[midenok]

These 2 tests are addressed in MDEV-10261.
innodb_onlinealter_encryption is addressed in MDEV-10099.