SGI: Development

Just as a note, std::bad_alloc is usually thrown for an out-of-memory condition, but it's also possible it's out of stack. Has anyone looked at the SP in the debugger when this occurs?

BetXen wrote: but I still have problems on my fuel (also 6.5.30). It is probably due to bad library paths and so on. I could identify pendant references to libgdk_pixbuf (now "ldd .../firefox'bin" gives no error anymore) and tried to renew .rqsfiles as ShadOfBlue suggested... without success for the moment. I always have the std::bad_alloc issue.

I would try to reinstall all nekoware packages listed in the release notes. I did not see this error so far. Memory should not be the problem, firefox3 starts with about 130mb and if you open very much tabs you may need up to 300mb. I write this message on my fuel currently, so it should run on your fuel without problems.

I got it installed and working on my Onyx2 today (the system hasn't been used since ~2007 so it took a lot of updating). Everything seems fine; the Javascript on twitter.com is borked, but that's probably a version thing rather than a bug.

canavan wrote: Your screenshot looks like you have not configured fontconf at all. Try adjusting things in /usr/nekoware/etc/fonts/local.conf and adding some more fonts

Thanks, canavan ! I do have fontconfig configged but your additions look great. I'm headed off to download your freetype upgrade now. The screenshots are deceiving, since I have to resize them. Here's a real one :


I wonder if this ever went anywhere ?

http://markmail.org/message/eee27p5al6xiwgtm

Just got done installing the latest revision ff3.0.19 on an Onyx350 with IP graphics (dual-pipe, 8P, 16GB).

Before installing I backed up my existing bookmarks and deleted the contents of ~/.mozilla/firefox/Cache.

Discovered a couple of issues - but those could easily be on my end:

One crash so far:

Code: Select all

|$(0x6dd)ALERT: Process [firefox-bin] 4512 generated trap, but has signal 10 held or ignored
....which is the same crash I'd get with neko_ff2 on a fairly regular basis.

I also noticed that ff3 seems to render some pages noticeably slower than ff2. As an example if I open any of the nekochan sub-forums, I can watch as ff3 renders a line for each topic in the forum one at a time, while that same process was almost instantaneous under ff2.

I didn't use the rulesoverride flag and provided ff3 with all of the dependencies it asked for. If I can provide any additional info (that might be more helpful) just let me know.

Update : I uninstalled ff2 and the issue with slow rendering went away. Since the uninstalling ff2 I also haven't been able to reproduce the trap error with ff3, so it looks like removing ff2 may have resolved the issues I experienced.

recondas wrote: Just got done installing the latest revision ff3.0.19 on an Onyx350 with IP graphics (dual-pipe, 8P, 16GB).

Do you still run the IR graphics ? Would be interesting to hear a comparison.

One crash so far:

Code: Select all

|$(0x6dd)ALERT: Process [firefox-bin] 4512 generated trap, but has signal 10 held or ignored

....which is the same crash I'd get with neko_ff2 on a fairly regular basis.

You'll get more, promise So far, FF3 crashes about as frequently for me as FF2. But the page rendering is superior. And the crashes have been addressed in other versions of FF so there is a potential upgrade path, where FF2 was a dead end.

Update : I uninstalled ff2 and the issue with slow rendering went away. Since the uninstalling ff2 I also haven't been able to reproduce the trap error with ff3, so it looks like removing ff2 may have resolved the issues I experienced.

<aol> me, too !</aol> I've never been lucky with having two versions of FF installed.

I still see where some websites will seemingly grab the display, so that changing tabs does not redraw the windows in other tabs, but again, that's a FF issue, not a porting issue.

If you have no-script, don't uninstall it. It puts some crap somewhere in the FF code that it doesn't remove. Every time I click rmb now there's an error code thrown. (I start it from a terminal at this time so that I can see what it's doing.) That wasn't there before I uninstalled no-script.

diegel wrote: It's in beta now: ftp://ftp.nekoware.de/beta/neko_firefox ... a3.tardist

now that's a great surprise. very cool stuff diegel
i'm still somewhat skeptical about the gcc faster than mips pro builds but firefox3 is different so i'll see soon enough.

this is a great plus for the irix community and it just feels good to see there's still active development for the "original" sgi gear going on!

foetz wrote: i'm still somewhat skeptical about the gcc faster than mips pro builds but firefox3 is different so i'll see soon enough.

You can try it. It will compile with the included patch with mips pro. The javascript engine will not run at all, but you can move the gcc compiled libmozjs.so to the mips pro build. I didn't test it very much, because for me it was slower than the gcc build and it was very unstable. With firefox 3.5 the libmozjs.so workaround will not work anymore, since it contains c++ code.

hamei wrote: Do you still run the IR graphics ? Would be interesting to hear a comparison.

Installed ff3 on the the Tezro, Octane and Onyx350 IR yesterday afternoon. Works equally well on all of 'em.

hamei wrote: You'll get more, promise So far, FF3 crashes about as frequently for me as FF2.

Knock on wood - ff3 hasn't crashed since I figured out the remove-ff2-first part.

hamei wrote: you have no-script, don't uninstall it. It puts some crap somewhere in the FF code that it doesn't remove. Every time I click rmb now there's an error code thrown. (I start it from a terminal at this time so that I can see what it's doing.) That wasn't there before I uninstalled no-script.

NS ( and the rest of the security add-ons I mentioned in the other ff3.0.19 thread ) are all up and running in ff3 w/o problem.

recondas wrote: Knock on wood - ff3 hasn't crashed since I figured out the remove-ff2-first part.

Go to http://www.sfgate.com Open about ten tabs. Right in the middle of a story about the damned ravens, crash.

Maybe it's for the better after all.

NS ( and the rest of the security add-ons I mentioned in the other ff3.0.19 thread ) are all up and running in ff3 w/o problem.

If you're not getting the

terminate called after throwing an instance of 'std::bad_alloc'
what(): std::bad_alloc

or any memory crashes, then cool. N-S was not a problem until I removed it. Then it left some junk behind somewhere.

If you don't crash at sfgate, we should compare ldd lists and library versions. I'd like to not crash too.

Or I could just avoid sfgate. It is kind of a birdcage liner. In the past couple of days they have printed two stories that were eighteen and fifteen years old. Let's have a hi-tech knowledge-based economy !

recondas wrote:

hamei wrote: Do you still run the IR graphics ? Would be interesting to hear a comparison.

Installed ff3 on the the Tezro, Octane and Onyx350 IR yesterday afternoon. Works equally well on all of 'em.

I'm running it on an IR3 system myself. No major issues.

hamei wrote: Go to http://www.sfgate.com Open about ten tabs. Right in the middle of a story about the damned ravens, crash.

If you're not getting the 'std::bad_alloc' or any memory crashes, then cool. N-S was not a problem until I removed it. Then it left some junk behind somewhere.

You might want to consider re-adding noscript - here's the aforementioned ten tabs of sfgate, plus a nekochan for good measure: Read a couple of the articles on sfgate and even followed a couple the links - didn't experience any crashes.

so i took this cutie for a ride and it was absolutely flawless. big thanks to all involved, i can imagine what a pita it was rigging it for irix.

as for the speed, page rendering did indeed seem faster than my last gtk1 firefox2 mips pro build although gtk2 always feels somewhat sluggish. that's the actual downside for me since i just never liked the looks of gtk2 not to mention that the dialogue windows are twice as slow as their gtk1 counterparts. obviously i never used gtk2 much so there might be config options to change a few thing that i don't know.


bottom line this is a kick ass contribution and a serious gain for any irix user

recondas wrote: You might want to consider re-adding noscript ... Read a couple of the articles on sfgate and even followed a couple the links - didn't experience any crashes.

Did you hold your nose ?

Since removing n-s hasn't helped, I probably will re-add it. It may help, but ... We have a slightly unusal situation here. Connections inside China are pretty fast but to overseas they are slow. The gfw slows things down and the backbone to the outside world is undersized. On top of that, facebook and tweeter and much of that crap does not get through, so the stoopid effing browser sits there and waits. And waits. And waits. And waits. Finally after ten minutes or so it gives up. The blithering idiots who coded this pile of crap never considered that maaaybe some day a site would be down.

I just ran through this again (twice.) It's repeatable :

( restore previous session )

The interesting part is, if I stick to old-fashioned websites (nekochan, graphics magick, personal sites) this is not a problem at all. It's when you go to the javascript-laden advertising-driven "rich user experience" modern sites-of-shit that all hell breaks loose.

Obvious solution : don't go there

Thanks for running a control test. Most likely I just have to work around this behavior.

I've toyed with the notion of using a binary editor to patch MIPSPro to add the c++ features it needs to stay relevant to code like that of ff18. But then I've toyed with the notion of defying gravity and faster-than-light travel as well...

And for what it's worth I'm here using the Mozilla 1.7.12 that came with my Octane, which is vintage September 2005...

hamei wrote: The interesting part is, if I stick to old-fashioned websites (nekochan, graphics magick, personal sites) this is not a problem at all. It's when you go to the javascript-laden advertising-driven "rich user experience" modern sites-of-shit that all hell breaks loose.

Yes, you can crash firefox with such kind of sites, but I did not see the errors you reported. I have seen a pango related error, I am quite sure this is a problem of the current pango package. I also have seen a bus error, but I was not able to reproduce it, so I can't debug it. There must be something different with your setup, are you using your firefox with a remote X11 server?

diegel wrote: I have seen a pango related error, I am quit sure this is a problem of the current pango package.

Agreed, we can put this one to bed. Maybe there are some problems with the current pango but the ones I was seeing came from an old test pango. The one in /current has not done that "warning : distorted display" thing.

So one down

I also have seen a bus error, but I was not able to reproduce it, so I can't debug it.

Same here, doesn't seem to be related to any particular situation or site. Except maybe sfgate but recondas did not have any problem there.

There must be something different with your setup, are you using your firefox with a remote X11 server?

No, pretty much the normal on-box local operation. It appears to be a combination of goofball sites with bad coding, the great firewall, our internet speed in general, China Telecom DNS which lies, and the waiting for facebook, tweeter, and other advertising which never shows up.

Some of this I just have to live with. Some of them are also apparent in the semi-current Firefox on Windows. Just a couple versions ago it would do things like freeze with 100% cpu. Forever. (Semi because they dropped Win2k a couple versions ago.) I would like to move the Firefox coders to China for about a year and force them to use their own product

Overall, it's a definite improvement. So I am not complaining, for sure. Thanks again !

edit : just timed it. http://www.tech-recipes.com Waiting for platrform.twitter.com ... three minutes twelve seconds !! Ridiculous !

hamei wrote: just timed it. http://www.tech-recipes.com Waiting for platrform.twitter.com ... three minutes twelve seconds !! Ridiculous !

So reload noscript and stop javascript from trying to jam all that extra crap thru that little hole in the GFW.

You know what Richard Nixon said; "It is...indeed...a great wall."

So I did a quick test with no problems so far. Google, Google News ('there is a script hanging' - well stop it then!) and Heise.de are displayed very well.
But as always browsing takes 100% of one CPU, so upgrading to R14k seems to be the only option for faster browsing (Web 2.0 WTF...!?)