[Rspamd-Users] Intermittent Unit Test Failures

[Steve Sturges (ststurge)]

On Apr 27, 2022, at 1:47 PM, dunc <a.16bit.sysop at gmail.com<mailto:a.16bit.sysop at gmail.com>> wrote:



On Mon, 25 Apr 2022, 18:17 Steve Sturges (ststurge), <ststurge at cisco.com<mailto:ststurge at cisco.com>> wrote:
Built from source within Alpine. It uses luajit.

On Apr 19, 2022, at 2:34 PM, dunc <a.16bit.sysop at gmail.com<mailto:a.16bit.sysop at gmail.com>> wrote:

Are you building it from source?
Are you using Lua or luajit?


On Thu, 31 Mar 2022, 17:36 Steve Sturges (ststurge) via Users, <users at lists.rspamd.com<mailto:users at lists.rspamd.com>> wrote:
I’m running rspamd 3.1 within an Alpine 3.15 container (Lua 5.1/LuaJIT 2.1) and have seen some intermittent failures with the
rspamd lua unit tests, in particular in lua/unit/lua_util.extract_specific_urls.lua.

Any ideas what could be causing this - it seems order of URLs in a lua table is not predictable?
I have not yet tried with rspamd 3.2.

From /usr/test within my container,

rspamd-test -p /rspamd/lua --verbose --debug-log

Across multiple runs, I see different results (output trimmed for simplicity):

Run 1:
641 tests 635 passed 809425 assertions 2 failed 0 errors 4 unassertive 0 pending

extract_specific_urls, backward compatibility case #3:
Failed asserting that
  (actual)   : {[1] = domain3.org<http://domain3.org/><http://domain3.org<http://domain3.org/>>, [2] = test.com<http://test.com/><http://test.com<http://test.com/>>}
 equals to
  (expected) : {[1] = domain4.co.net<http://domain4.co.net/><http://domain4.co.net<http://domain4.co.net/>>, [2] = test.com<http://test.com/><http://test.com<http://test.com/>>}
…

Run 2:
641 tests 637 passed 809517 assertions 0 failed 0 errors 4 unassertive 0 pending

Run 3:
641 tests 635 passed 809425 assertions 2 failed 0 errors 4 unassertive 0 pending

extract_specific_urls, backward compatibility case #3:
Failed asserting that
  (actual)   : {[1] = domain4.co.net<http://domain4.co.net/><http://domain4.co.net<http://domain4.co.net/>>, [2] = domain3.org<http://domain3.org/><http://domain3.org<http://domain3.org/>>}
 equals to
  (expected) : {[1] = domain4.co.net<http://domain4.co.net/><http://domain4.co.net<http://domain4.co.net/>>, [2] = test.com<http://test.com/><http://test.com<http://test.com/>>}
…

Run 4:
641 tests 635 passed 809483 assertions 2 failed 0 errors 4 unassertive 0 pending

extract_specific_urls, backward compatibility case #3:
Failed asserting that
  (actual)   : {[1] = test.com<http://test.com/><http://test.com<http://test.com/>>, [2] = domain3.org<http://domain3.org/><http://domain3.org<http://domain3.org/>>}
 equals to
  (expected) : {[1] = domain4.co.net<http://domain4.co.net/><http://domain4.co.net<http://domain4.co.net/>>, [2] = test.com<http://test.com/><http://test.com<http://test.com/>>}
…

Run 5:
641 tests 635 passed 809431 assertions 2 failed 0 errors 4 unassertive 0 pending

extract_specific_urls, backward compatibility case #3:
Failed asserting that
  (actual)   : {[1] = domain3.org<http://domain3.org/><http://domain3.org<http://domain3.org/>>, [2] = domain4.co.net<http://domain4.co.net/><http://domain4.co.net<http://domain4.co.net/>>}
 equals to
  (expected) : {[1] = domain4.co.net<http://domain4.co.net/><http://domain4.co.net<http://domain4.co.net/>>, [2] = test.com<http://test.com/><http://test.com<http://test.com/>>}
…

Run 6:
641 tests 637 passed 809517 assertions 0 failed 0 errors 4 unassertive 0 pending


--
Users mailing list
Users at lists.rspamd.com<mailto:Users at lists.rspamd.com>
https://lists.rspamd.com/mailman/listinfo/user<https://lists.rspamd.com/mailman/listinfo/users>

There is an open issue on alpine for luajit being compiled for sse4.2, and if you are using vectorscan it has recently had some merge requests to fix some random behaviour.

Could it be one of those?

Thanks… What I’m seeing definitely seems random in terms of a list of URLs that are extracted from a message and the order they appear.

I’ll take a look in a bit more detail, as it could be something related to that.  I doubt its an sse issue.