|
From: Bart V. A. <bva...@ac...> - 2010-02-15 13:23:44
|
On Mon, Feb 15, 2010 at 12:52 PM, Konstantin Serebryany <kon...@gm...> wrote: > On Mon, Feb 15, 2010 at 2:45 PM, Bart Van Assche <bva...@ac...> wrote: >> On Mon, Feb 15, 2010 at 12:11 PM, <dat...@go...> wrote: >>> Revision: 1655 >>> Author: kon...@gm... >>> Date: Mon Feb 15 03:10:36 2010 >>> Log: added ANNOTATE_BENIGN_RACE_SIZED; ANNOTATE_BENIGN_RACE_STATIC uses >>> *_SIZED variant now >>> http://code.google.com/p/data-race-test/source/detail?r=1655 >> [ ... ] >> >> Hello Konstantin, >> >> Have you already considered deprecating ANNOTATE_BENIGN_RACE(), since >> this annotation doesn't have precisely defined semantics ? > > Well, it does have precisely defined semantics now. :) > It is equivalent to ANNOTATE_BENIGN_RACE_SIZED(mem, 1, descr); As far as I know this is the first time it is specified how many bytes ANNOTATE_BENIGN_RACE() applies to. I'm not sure it is a good idea to choose a range of one byte, given that e.g. the following code is present in the TSAN unittests: int GLOB = 0; ... ANNOTATE_BENIGN_RACE(&GLOB, "test56. Use of ANNOTATE_BENIGN_RACE."); Bart. |