Bugs and memory hogs [LWN.net]

Bugs and memory hogs

Posted Jun 10, 2014 22:04 UTC (Tue) by NightMonkey (subscriber, #23051) [Link] (3 responses)

Just an aside - if actual software development actually matched the froth used to announce new releases, we'd have browsers so fast that they would browse for us before we even touch the input. ;)

Version 1 = faster
Version 2 = better stronger faster
Version 3 = better stronger faster less bugs
Version 4 = better stronger faster less bugs new company formed
Version 5 = better stronger faster less bugs leveraging synergies

etc.

:)

Of course, we do have a Turing test passing event recently, so maybe software has actually become faster and sentient? Signs point to "No"...

Bugs and memory hogs

Posted Jun 10, 2014 22:28 UTC (Tue) by coriordan (guest, #7544) [Link]

Of course, as soon as we detect sentience in computers, the computers will get to work on manipulating the evidence to make us quickly conclude the detection was just an error.

Version 6 = Sentient!?
Version 7 = Nope, our bad. Just too fast for the test suite.

Bugs and memory hogs

Posted Jun 11, 2014 13:00 UTC (Wed) by mathstuf (subscriber, #69389) [Link] (1 responses)

> we do have a Turing test passing event recently, so maybe software has actually become faster and sentient?

There have been reports that this is bad reporting at its finest. The only test was a 5 minute chat and they impersonated a teenage non-native speaker to waive away any language issues. Any report that calls it a "supercomputer" rather than a "chatbot" isn't doing their best reporting.

FWIW, I've seen Watson (from Jeopardy) in person and its language was much better than the stereotypical[1] English-as-a-second-language person (which it "spoke" rather than typing it out) and it bantered with the emcee, not just reacted with the questions necessary to the game.

[1]Certainly not all ESL speakers.

Bugs and memory hogs

Posted Jun 12, 2014 21:29 UTC (Thu) by h2 (guest, #27965) [Link]

What did it this time is Warwick's claim that the "Turing Test" - which measures ability of a machine to convincingly mimic a human while communicating with real humans in a blind test - had been passed at an event Warwick had organised and hosted. This had all the hallmarks of a Warwick stunt - you only had to look.

Warwick told the media that the landmark had been achieved using a "supercomputer" - when it fact it was a simple AI chatbot program running on a laptop. The chatbot's developer had tried and failed many times to convince humans it was human. This time, the academic luminaries chosen to judge the Test included a retired advertising being with no scientific background (now a Lib Dem peer) and, um … the TV actor and former shoemaker Robert Llewellyn, whose cybernetics qualifications consist of having played the neurotic robot Kryten in Red Dwarf.

the register. Before they pat themselves too much on the back for this 'expose', it's worth remembering they printed this story straight a few days ago. The comments had a few people who found the transcripts and concluded that only a total idiot could have fallen for this.

If I read it right, this 'supercomputer' was a basic laptop running some standard chatbot software.

Bugs and memory hogs

Posted Jun 10, 2014 22:53 UTC (Tue) by dlang (guest, #313) [Link] (19 responses)

are you sure it still has the memory problems? There's been a lot of work related to that over the last year, and I typically run a couple hundred tabs over a couple dozen windows and don't have it eating more and more memory over time.

Now, I don't use many extensions, so it could be that there is a problem with one of the extensions you are using.

I also haven't seen the 'scroll down at warp5' problem you describe

Bugs and memory hogs

Posted Jun 11, 2014 1:08 UTC (Wed) by dashesy (guest, #74652) [Link] (18 responses)

It was much better like 2 or 3 versions ago, recently I have to do a shift+F2 plus restart at least once a day, with 24GB of RAM. It becomes slow to response, to clicks, types, or scrolling for instances, it is very annoying.
I have at least one GMail open with Hangout for IM, and some gitlab pages, and it seems they kill FF. But I am sure 2 versions ago I did not see this much memory consumption.

Bugs and memory hogs

Posted Jun 11, 2014 2:04 UTC (Wed) by keeperofdakeys (guest, #82635) [Link] (4 responses)

What OS was this on? Firefox is still limited to 32-bit on Windows, which means it could never use lots of memory.

Bugs and memory hogs

Posted Jun 11, 2014 2:08 UTC (Wed) by dashesy (guest, #74652) [Link]

Fedora 20 with Cinnamon. I do not even dual boot Windows since the advent of Secure Boot, and Windows8 is a torture to touch.

Bugs and memory hogs

Posted Jun 12, 2014 18:44 UTC (Thu) by NightMonkey (subscriber, #23051) [Link] (2 responses)

But isn't each process limited to a 2GB 'box' of memory that can be exhausted, making that application's performance suffer should it near that limit?

Bugs and memory hogs

Posted Jun 14, 2014 2:49 UTC (Sat) by salimma (subscriber, #34460) [Link] (1 responses)

Not by default on 64-bit Linux, though you can set such limits yourself. I've certainly seen Firefox (pre-29) gobble up over 3 GB of RAM before. It got so bad I ran it through a script that'd terminate it if it reaches over 2.5G of RAM usage.

Bugs and memory hogs

Posted Jun 14, 2014 4:43 UTC (Sat) by NightMonkey (subscriber, #23051) [Link]

The post I was replying to referenced Win32. I knew about Linux/64-bit. :)

Bugs and memory hogs

Posted Jun 11, 2014 4:20 UTC (Wed) by drag (guest, #31333) [Link]

That is not normal...

My experiences with Firefox is that it's gotten generally better lately. The only problem I ran into was extremely slow start up speeds on my laptop, but I fixed that by clearing out the cache.

The only thing I can think of that comes close to your experience is that if I leave a LOT of tabs open over a weekend I'll come back to Firefox using quite a bit of swap. (older 4GB of RAM desktop) Otherwise it seems that while I am using it I only have to restart it every few days. This is certainly nothing new... it's been pretty much par for the course since Mozilla was a new project.

Maybe it helps that I don't install the Flash plugin.

Bugs and memory hogs

Posted Jun 11, 2014 6:19 UTC (Wed) by dlang (guest, #313) [Link] (9 responses)

also, check that it's actually memory that's causing things to slow down. I've been fighting a problem for a long time where FF starts chewing cpu in a tight spinloop, causing multi-second delays in responding to anything. I'm not the only one, but it's hard to duplicate and track down

Bugs and memory hogs

Posted Jun 11, 2014 9:20 UTC (Wed) by khim (subscriber, #9252) [Link] (5 responses)

I think it's related to connections in some form. When I try to use Firefox via GPRS the end result is invariably some stuck pages which chew 100% of the CPU. Even if they left alone for hours they continue to chew the CPU (and battery, of course). If you'll find the offending pages and do reload (or just close them) the problem goes away.

Bugs and memory hogs

Posted Jun 11, 2014 13:01 UTC (Wed) by mathstuf (subscriber, #69389) [Link] (4 responses)

I can confirm this. I've had tabs which were loading when my connection cut and Firefox will burn a core until you stop or get them reloaded. Luckily it is one per instance, not tab.

Bugs and memory hogs

Posted Jun 11, 2014 18:17 UTC (Wed) by dlang (guest, #313) [Link] (3 responses)

do you know of a way to identify which tab lost a connection and is spinning?

Bugs and memory hogs

Posted Jun 11, 2014 19:00 UTC (Wed) by mathstuf (subscriber, #69389) [Link] (2 responses)

There isn't a favicon on the tab, instead the spinny circle animation.

Bugs and memory hogs

Posted Jun 13, 2014 7:20 UTC (Fri) by kleptog (subscriber, #1183) [Link] (1 responses)

For some reason the Nagios pages at my work do this. The spinning never stops and the whole browser starts consuming more and more CPU until I kill the tab.

Bugs and memory hogs

Posted Jun 13, 2014 11:36 UTC (Fri) by mathstuf (subscriber, #69389) [Link]

It probably uses some JS to continually poll the server. I think it is possible to do "push" notifications, but I'm unsure. It should at least be on a timer to check.

Bugs and memory hogs

Posted Jun 20, 2014 2:05 UTC (Fri) by dashesy (guest, #74652) [Link]

I am now keeping an eye on it, and yes it seems it is more a CPU hike than memory, even though memory does increase (mostly windows objects and usually with Google domain).

Bugs and memory hogs

Posted Jul 18, 2014 23:09 UTC (Fri) by blujay (guest, #39961) [Link] (1 responses)

This may be the known issue with the session saver framework, which runs periodically and synchronously in the main loop, serializing to the sessionstore.js file, which can grow to several megs of JSON data. Work is ongoing to move it to a separate thread, and has been for years. One of these days...

Bugs and memory hogs

Posted Jul 18, 2014 23:45 UTC (Fri) by dlang (guest, #313) [Link]

that sounds like it would match my behavior very well. I've noticed that when sessionstore.js gets large the problem gets worse (not always, but if it gets >1m or so I have problems FAR more frequently)

do you have a link to that bug?

Bugs and memory hogs

Posted Jun 11, 2014 12:54 UTC (Wed) by wazoox (subscriber, #69624) [Link]

In my experience this was related to various extensions (I have 40 extensions at any moment). Generally FF works perfectly fine for me under Linux, but still gobbles RAM like there's no tomorrow under Mac OS X.

Bugs and memory hogs

Posted Jun 18, 2014 12:35 UTC (Wed) by jzbiciak (guest, #5246) [Link]

It's been like this for me for a long time. I've found that certain pages, if left open, seep memory. The about:memory window does have a "minimize memory usage" button that helps a little, but it's not a panacea.

For the memory-seeping pages, once their usage gets above a few hundred MB (and stays there after "minimize memory usage"), the browser gets unresponsive. Closing that tab often restores the browser's responsiveness.

In fact I tried it just now. I had a couple tabs open on "gocomics"—a comic that Bill Watterson snuck out of retirement to do a few panels for—that I hadn't closed, and it had climbed to the ~400MB range over the last week. This text-editing box had gotten very choppy, until I tabbed over and closed them. Now it's almost as responsive as if I had just restarted the browser. (SuperBrightLEDs is another site with this property; I reported that one to the Firefox team. The built-in PDF.js PDF viewer seems to be another.)

I suspect it had a bit of Javascript that was churning in the background, and whatever loop was churning was dragging the whole Firefox event loop down.

It's amazing. It's not the total memory size that Firefox is consuming that slows it down. It's one or two pages that locally seem to be churning through something continuously in the background. FWIW, this Firefox instance has been open nearly a month, and has a RSS around 2.4G. (Total VSZ around 7.5G, but partly because I experimented with upping some Javascript heap GC parameters for fun into the 5-6GB range; didn't really change responsiveness, but did change the memory footprint between pushes of the "minimize memory usage" button.)

I'm currently running 64-bit Firefox 29 on 64-bit Linux. I've had this "gradually slow to a crawl" behavior for quite a long time (going back more than a dozen versions), but only recently began experimenting with leaving pages open and monitoring them to see if I could identify a cause.

No idea if browser extensions are part of the equation. I do run a few, but mostly of the AdBlock/FlashBlock variety.

Bugs and memory hogs

Posted Jun 11, 2014 1:15 UTC (Wed) by roc (subscriber, #30627) [Link] (1 responses)

When you're consuming lots of memory, please file a Bugzilla bug with a report from about:memory.

Bugs and memory hogs

Posted Jun 11, 2014 1:16 UTC (Wed) by roc (subscriber, #30627) [Link]

Also, try reloading just individual apps and see if you can narrow down to a particular app. about:memory helps with that too. Sometimes Web apps leak and there's not much we can do about it...

Bugs and memory hogs

Posted Jun 11, 2014 3:51 UTC (Wed) by sanjoy (guest, #5026) [Link] (1 responses)

With a recent Debian testing update, I jumped from FF 24 to 29. This is on a 32-bit i386 system (T60, 3GB of RAM). FF 24 was a memory hog, growing over days to several GB, until it would get killed by the OOM killer. But FF 29 is a vast improvement. It has stayed at 1.4GB for several days now (0.5GB resident).

I don't use gmail or IM, so maybe those cause extra problems. But Mozilla seems to have fixed a lot of memory leaks.

Bugs and memory hogs

Posted Jun 11, 2014 10:25 UTC (Wed) by salimma (subscriber, #34460) [Link]

On my work laptop with 4GB of RAM, FFox 29 is probably the first release I can remember where I don't have to restart the browser after several hours (running tens of tabs).

Bugs and memory hogs

Posted Jun 11, 2014 10:57 UTC (Wed) by cesarb (subscriber, #6266) [Link]

If you use Adblock Plus, take a look at https://bugzilla.mozilla.org/show_bug.cgi?id=988266.