I've found a bug in OWB 1.20.

When I use Gmail in "enhanced" mode, the browser keeps eating up a little memory at the time until the whole RAM is gone. It fixes easily by closing Gmail.

Maybe something you can do, Fab? :)

It would be a big job for him to teach WebKit authors to code properly ;)

I have a question for Fab if he's reading this.

Are you planning on keeping WebKit as the engine for Odyssey or migrating to the Blink engine at some point?

Or are you just waiting and seeing what transpires?

> Are you planning on keeping WebKit as the engine for Odyssey or migrating
> to the Blink engine at some point?

I think I read that there's some intertwining in place (or at least planned) between Blink and the V8 JavaScript engine. Thing is that so far, V8 is only available for x86(-64), ARM and MIPS. There've been attempts at porting it to Power Architecture but none has come to pass so far.
(Yes, I know I'm not Fab but replied anyway ;-)

I don't think there would be any point in migrating to Blink, because as Andreas said it's tied to V8. It's basically google's customized webkit fork for Chrome. As a result of the fork Webkit will be smaller, because they are removing Google's cruft from it.

Quote:

:)

Ah yes I didn't think about that.

So if I understand correctly then, the JavaScriptCore engine currently used in Webkit has two versions. One for x86/64/arm that JIT compiles the JS and another for PPC et al that is purely an interpreter?

Or is there just one engine and it also JITs for PPC too?

If the former, what happens if/when Webkit drops their interpreter?

Just thinking out loud. :)

Just found this interesting project from an IBM employee whose full-time job is to get V8 working perfectly on PPC.

https://groups.google.com/forum/#!msg/v8-dev/06ReYVfFgGA/KO9Xr4SEBhQJ

So there could be a slim chance for a Blink based browser on MOS at some point. ;)

> So if I understand correctly then, the JavaScriptCore engine currently used in Webkit
> has two versions. One for x86/64/arm that JIT compiles the JS and another for PPC
> et al that is purely an interpreter?

Yes. You can read more from Fab there:

http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=37255&forum=46#697350
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=37255&forum=46#697354

> Or is there just one engine and it also JITs for PPC too?

No.

> If the former, what happens if/when Webkit drops their interpreter?

Given that they quite recently changed their interpreter to a new one, I don't think they'll drop it.

> Just found this interesting project from an IBM employee whose full-time job is to
> get V8 working perfectly on PPC.
> https://groups.google.com/forum/#!msg/v8-dev/06ReYVfFgGA/KO9Xr4SEBhQJ

Nice find, and sounds promising :-)

"July 23th 95% of the tests were passing. (with crankshaft!)"
https://github.com/andrewlow/v8ppc

Quote:

I want, I want, I want! :)

Quote:

Very cool, and about damn time

Quote:

Perhaps this project and all that Open POWER stuff in the press recently is the beginning of IBM trying to push PPC back towards the mainstream?

Then again, it's probably just so they can run node.js on V8 on POWER servers.

Quote:
Google is part of the Open POWER initiative, maybe there's some link between that.

On the other hand, i think it's nice that google doesn't touch webkit svn anymore. They introduced quite some mess that i'm happy to see gone. So going to blink doesn't look very appealing to me for now.

Implementing the llint ppc backend in webkit would be easier than JIT and would provide a 2-3x speedup already. And actually, the responsible for leopard webkit project also started to work on JIT, trying to integrate tenforfoux's one. It's already functional for YARR jit (regexp stuff), but for javascript, many trampolines still have to be written.

[ Edited by Fab 09.08.2013 - 10:50 ]

Scrub that.

Must have messed uip some settings somewhere.

[ Edited by boot_wb 09.08.2013 - 11:18 ]

Update:

> "July 23th 95% of the tests were passing. (with crankshaft!)"
> https://github.com/andrewlow/v8ppc

"August 9th 98% of the tests were passing. (with crankshaft!)"

Quote:

It's looking very promising but Fab has
made it clear he's sticking with WebKit, so unless he changes his mind or someone else starts a new browser based on Blink/V8 it doesn't look likely it will be coming our way anytime soon.

Does anyone know what GUI toolkit Chromium uses? Not that it makes much difference to us anyway. :)

> Then again, it's probably just so they can run node.js on V8 on POWER servers.

While it's about node.js, it's apparently not just about IBM POWER:

"If you're G4 Powerbook is running Linux - I think you _are_ the target market :)"
https://groups.google.com/forum/#!topic/nodejs/ivGMr-C_m1Y


Edit: The author plans to talk about his work 4 months from now:

https://www.eclipsecon.org/na2014/session/nodejs-powerpc-story-about-porting-nodejs-v8
https://www.eclipsecon.org/na2014/sites/default/files/slides/Andrew%20Low%20-%20PowerNode.pdf


Edit2: Final version added:

https://www.eclipsecon.org/na2014/sites/default/files/slides/Node.js%20on%20PowerPC.pdf
http://lowtek.ca/talks/2014-03-19-EclipseCon/

[ Edited by Andreas_Wolf 20.03.2014 - 20:13 ]

Quote:

GMail and enhanced mode has nothing to do with it. I get it even if I don't use Gmail. The RAM just keep draining out by itself even if I am at kitchen drinking coffee.

Quote:

As said, it's incompetent webkit (mainly javascript core) developers' fault. They just leak the mem without any care. It happens with all WebKit based browsers, but on other systems there's virtual mem to lean on. And as someone once said, Apple fixed this issue by going to 64 bit to be able use more ram ;)

Fab has already done much for it. He has made the browser quit and release the leaked mem if you ran out of it, instead to just crash and burn, but he can't change all mem allocations etc in the whole code himself.. it's just that really bad coding manners they have today on big world where they don't have to care about resources.

But if you want to save mem, disable javascript on OWB and enable it per site if you need. Or just restart the browser occasionally.

@ jPV

OWB has the same bug. 100megs are free but OWB says it cannot allocate 100kB. That doesn't make sense but it happens at least once a day.

Incompetence or not: best way to get rid of Webkit bugs is to to tell their bugtracker about it.

Quote:

Myeah that has been my strategy.. its just a bit frustrating to relaunch OWB every single morning during the 30 mins of my morning coffee

Here is a page I can't login to. I simply cannot push the login button ("Logga in"). And an interesting note appears at the bottom of OWB that probably tells where the problem lies.

http://bokabostad.stockholm.se

While using OWB for a while I notice that a lot of RAM becomes unavailable, from opening TABs, and eventually closing them all, but one.

So starting off with just about 900MB of RAM it gets to a point that there is basically no RAM left, even though OWB is the only application that has been opened, and at that point there is only one window/TAB left open, but I can't find a way to free back up the rest of the RAM in my Mac Mini that should be available.

Is there some utility that can release this unavailable RAM?

I tried 'avail flush' but that barely releases any of the RAM.

As soon as I close OWB the RAM frees itself back up, and then I have just over 900MB of RAM available again, so somewhere within the program it isn't releasing what it's not actually apparently using - is this a known bug?

> Is there some utility that can release this unavailable RAM?

Freeing RAM can only be done by the program occupying it.

> I tried 'avail flush' but that barely releases any of the RAM.

Yes, this is expected.

> somewhere within the program it isn't releasing what it's not actually apparently
> using - is this a known bug?

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=8977&forum=3