[Server] one_result_per_user_per_wu potentially broken

[polymetric]

Describe the bug
The server configuration option one_result_per_user_per_wu seems to not function as expected. I've seen 3 separate BOINC servers that have this option enabled and some users get two tasks from the same workunit. If the one_result_per_host_per_wu option is not enabled, a single host can get both tasks from a workunit:

I spoke with the SiDock@home admin and, at the time of that log entry, one_result_per_user_per_wu was enabled but one_result_per_host_per_wu was not.

Affected Projects

• Minecraft@Home
• SiDock@home
• my own private BOINC instance

Steps To Reproduce

1. Enable the rule one_result_per_user_per_wu in config.xml.
2. Wait for some new tasks to be distributed.
3. Go to the ops panel, click 'Workunits' under 'Browse database', then paste this into the 'additional clauses' box: id in (select r.workunitid from result r join result r2 on r.workunitid = r2.workunitid and r.hostid <> r2.hostid and r.userid = r2.userid)
4. Alternatively, in a clean BOINC server test environment, simply attach one host and watch as it gets both replications.

Expected behavior
The query should return no results - hosts should not get two results from the same workunit.

Screenshots
my BOINC instance:

Minecraft@Home:

SiDock@home:

System Information

• OS: Debian 10 Buster (used by my private BOINC server)
• BOINC Version: 1.2.0 (https://github.com/BOINC/boinc/tree/server_release/1/1.2) (version used by both my server and Minecraft@Home)
• BOINC Version: 1.1.0 (version used by SiDock@home)

Additional Context
Minecraft@Home config.xml: https://github.com/minecrafthome/minecrafthome/blob/master/images/makeproject/project/config.xml#L35

Feel free to ask for additional clarification if I missed anything!

[davidpanderson]

Do you happen to know if both results were returned in a single scheduler request?

[polymetric]

That I don't know, sorry.

[AenBleidd]

@lfield, is there any chance you could try to take a look at this issue?

[lfield]

I can not reproduce this issue on LHC@home dev. I notice that those two projects are not running the latest code.