[Swan-commit] Changes to ref refs/heads/main

Andrew Cagney cagney at vault.libreswan.fi
Fri Oct 9 15:20:25 UTC 2020


New commits:
commit 4091458afdf6c05aeac367296109931308df711d
Author: Andrew Cagney <cagney at gnu.org>
Date:   Thu Oct 8 16:09:51 2020 -0400

    kvm: add post-mortem.sh - silently run on all the guests after the test scripts
    
    This provides a mechanism for addressing two seemingly conflicting
    requirements:
    
      - ensure that all tests stop pluto and check for leaks et.al.
    
      - skip stopping pluto (...) (but still get a valid result) when
        running only a single test
    
    Once all the test's scripts have run, the script post-mortem.sh is run
    on the guests.  By default the script is silent - since no output
    appears in the sanitized *.console.txt it has no affect on test
    results.  Only when the script barfs (i.e., non-zero exit) does its
    output appear (see post-mortem.sed).
    
    Currently post-mortem.sh does little.  The next two steps are to add
    all the boiler-plate checks such as stopping pluto to post-mortem.sh,
    and then remove those checks from final.sh.
    
    The script is run with one exceptions:
    
       WHEN THERE IS ONLY ONE TEST THE SCRIPT IS SKIPPED
    
    for instance either of:
    
       kvmrunner.py testing/pluto/impair-12-ikev2-transform-none
       make kvm-test KVM_TESTS=testing/pluto/impair-12-ikev2-transform-none
    
    will not run post-mortem.sh (use --run-post-mortem true to override
    this).  Since running post-mortem.sh is silent this can't be seen and
    the test result is valid.  Since post-mortem.sh stops pluto and that
    is skipped pluto is left running.
    
    One wart is the silent but deadly sanitizer cut-postfinal.sed.  If
    final.sh emits the '=== end ===' marker then all output following,
    which would include a post-mortem.sh barf, is lost.  A possible
    work-around is to add a check for a post-mortem barf in
    *.console.verbose.txt to fab/post.py.



More information about the Swan-commit mailing list