Hacker News Reader

Comment by nh2 | Hacker News Reader

Firefox has significant performance and correctness problems.

* Multi-second UI jank/freeze (https://bugzilla.mozilla.org/show_bug.cgi?id=1546847)

* HTTP2 is broken, infinite-stalls and internal errors surface if you upload enough files (https://bugzilla.mozilla.org/show_bug.cgi?id=1540574)

* WebGL is very slow on Linux (https://bugzilla.mozilla.org/show_bug.cgi?id=1010527)

* High CPU on Linux even when in the background, possibly due to frequent polling (https://bugzilla.mozilla.org/show_bug.cgi?id=508427)

These are just a few issues that I'm directly involved in and suffer from every day, and into which I've sunk over 100 hours profiling and reading code of, simply because they make it an absolute pain to use.

View on HN

> High CPU on Linux

Provided this is anecdotal evidence, in my experience it's the opposite. I gained an enormous amount of performance-related comfort after switching to Linux Firefox from Windows Firefox/Chrome (where a JS-heavy web site that I use frequently would cause memory overflow and would stall the laptop). I've not experienced any problems with CPU usage since. I've not experienced the other problems either, though I don't use WebGL. You make it sound like Firefox is heavily broken for everyone. That's not the case.

I can't really speak to the Windows (FF/Chrome) -> Linux (FF) switch, but I have seen so many Windows setups that 100%-CPU-loop for unknown (likely browser unrelated) reasons that I would not be surprised if people considered switching to what I experience now a major improvement :)

> You make it sound like Firefox is heavily broken for everyone. That's not the case.

I didn't say it's unusable. But these things matter. If you get constantly annoyed by such things and you know that with Chromium this annoyance would stop, it is no surprise people use Chromium a lot and claim it's "faster". If we want them to be happy Firefox users, this has to be fixed.

Thank you for your work in helping profile FF's issues.

The first issue was really enjoyable to read. A high quality bug report and the devs giving feedback.

Why did you reference that last issue? I haven't had that CPU usage issue since Quantum was released.

> Why did you reference that last issue?

Because it's still happening, also on Quantum. I still see the 5-15% CPU usage on idle. If you strace Firefox, you'll see the syscall spam, even when the browser is apparently fully idle. This consumes a lot of battery power because it prevents the CPU going to sleep, even when you're looking at a still screen in Firefox (compare to gnome-terminal as mentioned in the bug report, which stops spamming as soon as nothing moves).

My laptop lasts only half as long on battery when Firefox is on and showing a still screen.

I can confirm the strace messages and FF's 15% erratic usage on a dual core i3 Skylake laptop.

I guess I never noticed it compared to how FF used to chug after long use and Google Chrome's same usage of around 30-40% of my cores.

I've certainly had something equivalent to the first one in Chrome frequently, seemingly when it's waiting for an unresponsive child process, often when closing a tab. It won't lock up as the UI seems to be separated from the other processes well enough to allow it to keep rendering loading animations etc. However every part of the browser will stop functioning including showing any tab content or opening any menu, for 10s of seconds at a time.

I've also had Chrome destroy my browser profile completely many times. It will start up and say something like 'your profile was unrecoverably corrupted and has been deleted'

Both are very complex pieces of software with lots of bugs.

I can confirm both.

Often, closing a tab in Chromium simply does nothing.

Chromium has lost my tabs multiple times (not by profile corruption, but not restoring them on startup). Firefox hasn't lost my tabs for over 10 years -- one big reason I use it (it did lose them ~5 times but I could always restore them from the backup file in the profile directory).

I wonder if Chrome is using Sqlite under the hood, I know Electron apps will lose their data entirely if any bit of the database is corrupt (Sqlite doesn't handle inconsistent states well :c).

On a slightly different tangent, you likely have a failing disk or RAM.

Yep, certainly possible. I think Firefox avoids this by keeping a few complete profile backups to fall back on.

Firefox also has bad input lag compared to chrome, everywhere from the address bar to basic textareas.

I've filed https://bugzilla.mozilla.org/show_bug.cgi?id=1408699 quite a while ago with some measurements. Sadly, no progress so far.

Can confirm. For me also sometimes input "buffers up", not showing up for over a second, and then appearing at once. I attribute most of that to the first bug in my list for now, as it freezes everything.

I’m guessing most people don’t suffer from the first issue because they don’t have 800 tabs open? The bug reports were very nicely done, however.

I’d recommend using bookmarks instead of keeping tabs open and unloaded, but I do think that insane numbers of tabs should be something that Firefox can handle.

> 800 tabs

As you commented, most of them are unloaded and only 10-20 are actually "real tabs". Have also tested with less by now, same problem.

On Chrome on Android I have more tabs, 16x less RAM and 50x less CPU power, and it doesn't jank at all, so this seems solvable.

I use OneTab[1] for Firefox for this. Bookmarks do have the advantage of being integrated to sync though.

[1] https://www.one-tab.com/

I've had a few instances of firefox on Mac and Linux starting to spin up the fans, and show high CPU usage, and I've had trouble pinning down which tab was causing it.

Is about:performance the best place to do this or is there something like Chrome's task manager that shows CPU and memory usage per page, and per extension?

Reply on news.ycombinator.com