Uploaded image for project: 'WiredTiger'
  1. WiredTiger
  2. WT-8986

Investigate clean up of posix_configure_flags in evergreen.yml

    • Type: Icon: Improvement Improvement
    • Resolution: Duplicate
    • Priority: Icon: Minor - P4 Minor - P4
    • None
    • Affects Version/s: None
    • Component/s: None

      Summary

      Split off from WT-8958. Original text:

      """
      The stress test variants (x86 and ARM) pass posix_configure_flags but the underlying tasks very likely ignore them and build their own version of WiredTiger. This causes confusion and could mislead us into making a change to the posix flags expecting a change to the build. Find a solution that the flags are not ignored, or remove them altogether so that we are not misled.
      """

      We currently set posix_configure_flags in around 40 locations in evergreen.yml.
      If we limit our scope to just the tests in the  Ubuntu 20.04 Stress tests (ARM64) buildvariant, posix_configure_flags is either:
      1. Set at the buildvariant level (csuite-wt8963-insert-stress-test)
      2. Overridden in an intermediate task (race-condition-stress-sanitizer-test), or
      3. Ignored entirely (format-abort-recovery-stress-test which calls compile wiredtiger with builtins)

      For consistency only one of these options should be used across all tests.

      Making the above change will touch most, if not all, tests in evergreen.yml so we need to determine what impacts this will have. In particular if this change results in new compile flags being used in existing tests and if the new flags change the behaviour of the test.

      Motivation

      • Does this affect any team outside of WT?
        No
      • How likely is it that this use case or problem will occur?
        Unlikely, this is mostly to reduce complexity/improve readability of our evergreen file
      • Is this issue urgent?
        No

      Acceptance Criteria (Definition of Done)

      This ticket should be a timeboxed investigation to determine the complexity of the proposed change and what impacts it will have. On completion we can decide whether to make the change or leave evergreen as is.

      Suggested Solution
      Some questions to consider:

      • If a test currently ignores the buildvariant posix_configure_flags what would be the impact of using these buildvariant flags? Where do they differ across buildvariants?
      • Tests like checkpoint-filetypes-test want to override a single compile flag. Is this possible without overriding the entire posix_configure_flags variable?
      • Are there any compile flags unique to specific buildvariants?

            Assignee:
            mick.graham@mongodb.com Mick Graham
            Reporter:
            andrew.morton@mongodb.com Andrew Morton
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated:
              Resolved: