Given the threads last week with Pino's claim that we had a memory leak, and Markus' counter-claim that the proposed patch was reverting an earlier double-free fix, I ended up enhancing the testsuite to prove under valgrind whether we indeed have a problem. The good news is that the existing visitor code is correct, but having more tests will help prevent future regression attempts :)
Eric Blake (2): tests: Simplify expected error checking for qmp output tests: Enhance qmp output to cover partial visit tests/test-qmp-output-visitor.c | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-) -- 2.7.4
