• unknown's avatar
    push for trnman review · c2872baf
    unknown authored
    (lockmanager still fails unit tests)
    
    
    BitKeeper/deleted/.del-Makefile.am~4375ae3d4de2bdf0:
      Delete: unittest/maria/Makefile.am
    configure.in:
      silence up configure warnings, don't generate unittest/maria/Makefile
    include/atomic/nolock.h:
      s/LOCK/LOCK_prefix/
    include/atomic/x86-gcc.h:
      s/LOCK/LOCK_prefix/
    include/atomic/x86-msvc.h:
      s/LOCK/LOCK_prefix/
    include/lf.h:
      pin asserts, renames
    include/my_atomic.h:
      move cleanup
    include/my_bit.h:
      s/uint/uint32/
    mysys/lf_dynarray.c:
      style fixes, split for() in two, remove if()s
    mysys/lf_hash.c:
      renames, minor fixes
    mysys/my_atomic.c:
      run-time assert -> compile-time assert
    storage/maria/Makefile.am:
      lockman here
    storage/maria/unittest/Makefile.am:
      new unit tests
    storage/maria/unittest/trnman-t.c:
      lots of changes
    storage/maria/lockman.c:
      many changes:
      second meaning of "blocker"
      portability: s/gettimeofday/my_getsystime/
      move mutex/cond out of LOCK_OWNER - it creates a race condition
      that will be fixed in a separate changeset
      increment lm->count for every element, not only for distinct ones -
      because we cannot decrease it for distinct elements only :(
    storage/maria/lockman.h:
      move mutex/cond out of LOCK_OWNER
    storage/maria/trnman.c:
      move mutex/cond out of LOCK_OWNER
      atomic-ops to access short_trid_to_trn[]
    storage/maria/trnman.h:
      move mutex/cond out of LOCK_OWNER
    storage/maria/unittest/lockman-t.c:
      unit stress test
    c2872baf
lockman-t.c 6.83 KB