Avoid valgrinding unit tests that intentionally leak memory,...

Document how we avoid running leaky tests under Purify and Valgrind,
and copy Purify's list of leaky net tests for Valgrind's benefit.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=12042

[dank, 2009-03-17 23:08:53]

Better than adding specific suppressions for each such test, IMHO

[Dean McNamee, 2009-03-18 11:05:52]

Looks ok to me, but you should run it past Erik Kay and Ricardo.

On 2009/03/17 23:08:53, dank wrote:
> Better than adding specific suppressions for each such test, IMHO

[dank, 2009-03-18 15:20:21]

Hi Erik,
Dean and I are trying to get the valgrind builder green.
I'd like to use a naming convention to tag tests that are
expected to leak memory; that will make it easier to 
filter out those warnings.
Look ok to you?
Thanks,
Dan

[dank, 2009-03-18 23:03:37]

Changed to follow the established pattern you pointed out,
and adds comments everywhere so it's easier to
figure out what's going on.  How's it look now?

[Erik does not do reviews, 2009-03-18 23:32:54]

LGTM.  Agree with you that we should have a shared filter file for common
exceptions (although likely also have a file that's specific to each).  Maybe
file a cleanup bug about that?

http://codereview.chromium.org/42309/diff/6/1005
File net/data/valgrind/net_unittests.gtest.txt (right):

http://codereview.chromium.org/42309/diff/6/1005#newcode1
Line 1: # These tests leak data intentionally, so are not appropriate for
Valgrind tests.
80 cols

[dank, 2009-03-18 23:42:42]

On 2009/03/18 23:32:54, Erik Kay wrote:
> LGTM.  Agree with you that we should have a shared filter file for common
> exceptions (although likely also have a file that's specific to each).  Maybe
> file a cleanup bug about that?

Filed http://code.google.com/p/chromium/issues/detail?id=8951