Http cache: fix a bug that prevented the infinite cache simulation from saving

Http cache: fix a bug that prevented the infinite cache simulation from saving
large data sets.

Also, go back to a limit of 200k entries, to monitor things, and increase the
version number to have clean stats after this change.

BUG=147383
TEST=none
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=170548

[tburkard, 2012-11-30 19:00:54]

lgtm

https://codereview.chromium.org/11415222/diff/5001/net/http/infinite_cache.cc
File net/http/infinite_cache.cc (right):

https://codereview.chromium.org/11415222/diff/5001/net/http/infinite_cache.cc...
net/http/infinite_cache.cc:88: int kMaxNumEntries = 200 * 1000;
Why reduce the size?

https://codereview.chromium.org/11415222/diff/5001/net/http/infinite_cache_un...
File net/http/infinite_cache_unittest.cc (right):

https://codereview.chromium.org/11415222/diff/5001/net/http/infinite_cache_un...
net/http/infinite_cache_unittest.cc:250: #if 0
How about not using #if 0, but making it Disabled, so that one can still run it
manually if needed?

[rvargas (doing something else), 2012-11-30 19:12:23]

Thanks.

https://codereview.chromium.org/11415222/diff/5001/net/http/infinite_cache.cc
File net/http/infinite_cache.cc (right):

https://codereview.chromium.org/11415222/diff/5001/net/http/infinite_cache.cc...
net/http/infinite_cache.cc:88: int kMaxNumEntries = 200 * 1000;
On 2012/11/30 19:00:54, tburkard wrote:
> Why reduce the size?

When I increased the size I was under the assumption that people were free to
reach the limit... knowing that most people have not really had a chance of
getting more than 100k entries means that we should be cautious again and see if
problems arise.

https://codereview.chromium.org/11415222/diff/5001/net/http/infinite_cache_un...
File net/http/infinite_cache_unittest.cc (right):

https://codereview.chromium.org/11415222/diff/5001/net/http/infinite_cache_un...
net/http/infinite_cache_unittest.cc:250: #if 0
On 2012/11/30 19:00:54, tburkard wrote:
> How about not using #if 0, but making it Disabled, so that one can still run
it
> manually if needed?

I don't care that much either way, but we have regular cycles of "what are we
doing with disabled tests, we should delete them or fix them" (which is 100%
valid concern) so I wanted to be explicit about this being intentional and not a
bug (plus less noise for everybody with "you have n disabled tests"). Running it
is a matter of removing two lines and a few seconds recompiling.

Sounds reasonable?