Re: Firefox 31.8.0: Use of Add-On "HTTPS Everywhere 5.1.2" causes 99% CPU utilization

Andreas Buchinger wrote:
> Dave Yeo schrieb:
>> Tom Brown wrote:
>>> Lars,
> .....
>>> While running either SM or FF, typically in cnn.com or
>>> news.google.com, ONE of the cpus will go to 99.9%, the status line at
>>> the bottom of the screen will display "waiting for" some site such as
>>> doubleclick.net or another ad site, and the whole system will hang. At
>>> this point, it is difficult to bring up the window list using Ctrl-Esc
>>> ahthough it can usually be done eventually. If I wait long enough, I
>>> might regain control. Sometimes, I have to C-A-D and kill the browser.
>>> I can often run another OS/2 app, such as solitaire, PMView, etc. to
>>> pass the time while waiting for the browser to get over it's choking
>>> fit.
>>>
>>> This sounds to me like a different situation from yours, but I would
>>> really like to find a solution.
>>>
>
> I do have similar problems - waiting for some site for minutes (?) with
> one of the latest SM from Dave. But do not see the 99% CPU load. Only SM
> is not responding.

I see similar I think, SeaMonkey doesn't load pages at times though the 
browser is still responsive

>
>> .....
>> My guess is a subtle JavaScript bug[s]. I use NoScript to block as many
>> scripts as I can get away with and have a fairly stable browser.
>> The ads are often compromised as well so at least in Toms case, it may
>> well be a script trying to take advantage of a security hole. While
>> often they won't work on our platform, they can still have negative effects.
>> Dave
>
> I use NoScript too. But in contrary my guess is some overload on the
> TCP/IP stack which delays address resolving (?). Probably triggered by
> JavaScripts which try to load so much useless stuff from so many sites.
>
> Sometimes when I start FF in parallel and try to load the same site this
> unblocks the situation (or maybe the hang timed out). Usually FF
> displays the same site slowly but it came up long before when SM is
> still waiting for some content. But after FF has loaded it, SM
> eventually get it too.

This I think is a different problem from the 100% CPU load triggered by 
JavaScript. Seems to be problems with sockets, perhaps not getting 
released. Sometimes restarting the browser helps, sometimes flushing the 
routing tables and sometimes only a reboot.

>
> IIRC even 'ping google.com' times out in such situations which let me
> think of a more fundamental problem than JavaScrips.

Try running \mptn\bin\setup.cmd. Sometimes fixes it here.
What is strange is the other computers on the network, all going through 
my dial-up connection keep working fine.

>
> One more observation is the newsreader seem to block whole SM sometimes
> too. I think especially when it tries to fetch news from sites hard to
> connect or temporary shut down (openwatcom.org, ecomstation.com). The
> browser do not get anything until some magic time out.

I get similar on Thunderbird, where all my mail is. There are a couple 
of design problems, too much running on the main thread and the mork 
database, which is also a flat, single threaded beast.
I do find that defragging (profiles have their own partition here. 
Delete all and restore from backup) helps quite a bit.

>
> Even worse was the situation when I had installed the calendar.
> Synchronizing with google/calendar (guess) let whole SM being unresponsive.
>
>>From all of this my guess is some sort of overload of our underlying
> system (stack? address resolution?). After some time out SM usually
> works again. Triggered maybe by JavaScrips (images.amazon.com, gmx.net).
>
> Only wanted to share my findings. But it is very hard for me to debug
> this better cause some of my problems maybe caused by my lousy ADSL
> connection. Sometime I have a Win machine running at the same time (not
> very often). And SM on Win usually does not show such problems. So for
> now I'm pretty sure there is something wrong with our port/system.
>

I think both. Old stack that could use some love and our port really 
could use a couple of full time programmers. My source tree is well over 
2 GBs, including the revision control stuff. It's a lot of code, which 
has been changing a lot and the changes will just keep speeding up. Like 
Cameron, it's hard to imagine keeping up and seems useless as they 
remove basic stuff like FTP.
Dave