[tor-qa] Tor Browser 4.5a5 is ready for testing

[Katya Titov]

Katya Titov:
> Katya Titov:
>> Katya Titov:
>>> Mike Perry:
>>>> Tor Browser 4.5a5 is available at:
>>>> https://people.torproject.org/~mikeperry/builds/4.5a5-build6/
>>>> 
>>>> This release will be the last alpha in the 4.5 series. We hope to
>>>> release 4.5-stable in mid-April.
>>> 
>>> This is my first look at the 4.5 series ...
>> 
>> [...]
>> 
>>> Other regular tests all seem OK and it works with Torsocks. Nice
>>> job!
>> 
>> Hmmm.
>> 
>> Firefox seems to be using quite a bit of CPU. CPU usage is
>> consistently 10-20% and there are about 100-200 more context
>> switches per second than when using TBB 4.0.6. vmstat output
>> attached. FF version is 31.6.0 in both, OS is Debian 7.8, processor
>> is Intel(R) Pentium(R) CPU B950 @ 2.10GHz, system has 4GB of RAM
>> and generally doesn't swap.
>> 
>> Nothing else was happening on the box at the time, and stats were
>> generated a few minutes after TBB had started in each case. I
>> haven't used earlier versions in the 4.5 series so I can't compare
>> this to older releases.
>> 
>> Please let me know what other information would be helpful.
> 
> This was consistently reproducible for most of the last 24 hours but
> has now stopped. When it stopped I noted that HTTPS-Everywhere and
> NoScript had been updated, and a restart confirmed the new versions
> (5.0.1 and 2.6.9.20 respectively). Maybe the update was the cause of
> the activity? Is this auto-updating of add-ons expected?

Still trying to narrow this down ... doesn't appear to be related to
updates (although I'm still interested in knowing whether automatic
updates are expected behaviour).

When the elevated CPU usage occurs the bottom of the FF window is
'pulsing'. I can't think of another way to describe it. The effect is
similar to the change in window display when the window comes into or
goes out of focus, but only affects the bottom of the window and not
the other sides. It is currently reproducible for me by attempting to
maximise the window. It happens at other times as well, although I
cannot find a consistent way to reproduce it.

The effect can be stopped by changing the focus of the FF window;
however, if I switch to another workspace and then back then this does
not work: I then need to stop/start TBB or attempt to maximise again
and change focus.

I use Xfce for my Window manager (stock Debian 7.8) and typically use
'focus follows mouse'. I have changed to 'click to focus' with no
change in result.

I can't find any relevant logs which may indicate what the issue is.

-- 
kat