REPRODUCIBILITY (% or how often): 100
BUILD ID = OS VERSION (Settings > About product): 4.3.0
HARDWARE (XA2, X10, X10 II, …): X, XA2
UI LANGUAGE:en
REGRESSION: (compared to previous public release: Yes, No, ?): yes
DESCRIPTION:
After some time of usage the browser (which starts really nice, fast and responsive) gets slow, sluggish and unresponsive.
PRECONDITIONS:
use the browser for some time and also write
STEPS TO REPRODUCE:
use browser
write text (wait seconds for it to appear)
try to scroll (touch not recognized, scroll stops for seconds)
EXPECTED RESULT:
browser keeps being responsive as it was on a fresh start
ACTUAL RESULT:
browser becomes.slow
ADDITIONAL INFORMATION:
(Please ALWAYS attach relevant data such as logs, screenshots, etc…)
this can be work-arounded for some time (until it happens again) by restarting the browser
Can you give us examples of websites that you visit? This isn’t reproducible to me, but smells like something wrong around JS/CSS. Nevertheless, I think a newer engine is coming soon™.
(But websites examples is very valuable)
I wonder if WebView performance is similar. To test the limits of WebView I built (using Alex Crists scaler):
That thing is a react native app (pared down, but…) and it becomes ever less responsive on input. It definitely is JS in this case. I’m going to compare this to running Scaler in the sfos browser.
It becomes sluggish in WebView relatively quickly. But, I also have the problem in the browser that it still won’t load zeit.de.
I wasn’t sure if it’s a different topic in the case of tcp/ip timeouts? The reason I bring that into the discussion is that I noticed that some slow (lots of tcp connections) sites load even more slowly in the 4.3 browser than in the 3.4 browser. It’s not evidence yet.
I do have a test case or two to check timings. I noticed that SailHN (a hacker news client in C++) doesn’t finish loading comments on threads in 4.3. Works in 3.4. Checking against another hackernews client (using python), I note that the comment threads also load on 4.3. Why would python fare better than c? My plan is to add some timing debug code to SailHN to see how the connections fare.
On the javascript side, I just tested the alexcrist scaler site directly in the sfos on 3.4 and 4.3. The 3.4 rendering and response to input is BETTER on 3.4 browser than either my 4.3 sfos browser OR my wrapper app.
Now that seems odd. But has nothing to do with TCP
It seems form elements are being repeatedly check with FormAssistant.js when it doesn’t seem they should be. As I understand it, FormAssistant is there to check if the Form Field that is in focus is a password form Username field and should not fire on every character input to ‘any’ form field. But it does.
yeah, I’ve noticed the same behaviour. I’m pretty sure it’s the javascript interpreter. If I just surf static html pages (I have a couple of sites like that), it’s always fine. Really minimal js sites also ‘never’ have this symptom. Anything with react js is slow… stops for a few seconds, recovers, etc.
