Just Passing Through

The latest Opera Browser on Debian Lenny AMD64: I don't write code, but I am certain there is something fundamentally wrong with an application when internal errors can take down the X server on a Linux box.

I don't file bug reports these days because it changes nothing. However, I'm not a happy camper with my Opera browser. It's not as bad as it has been in the past, but at this point I can't imagine this should still be happening. I visited a site which popped up several messages about "Opera does not support that function." Okay, so there is some poorly implemented Javascript here. That describes about half the websites on the entire Net.

Then the browser goes into a race condition and paints itself permanently on the GUI. I wait a bit, but it's still going, so it's not something Opera can resolve. Meanwhile, as I switch to another desktop to kill it, Opera paints itself on the screen there, too. Still, I manage to kill the process. Not good enough. Opera still paints itself on every desktop. It has damaged the X server itself.

It requires not just a restart of X, but I actually have to reboot to kill the child processes. That may be okay for Windows, but in Linux it means Opera programmers are ignoring the built-in mechanisms in Linux for preventing such things. Not a single other application I've ever used on Linux does such a thing.

I dunno, but maybe some day Opera will figure out how to code for Linux. They aren't there yet.

I think I finally understand why Opera works so poorly on Linux: It's really not a native port. That is, it's the only browser for Linux which requires its own personal wrapper for all the plugins. That's because Opera isn't really designed to use them directly, they way other made-for-Linux browsers do. Thus, this wrapper is the weakest link in the chain, and is the part which keeps going wonky.

Whenever things don't go quite perfect with, say, a Flash video, the wrapper enters what we call a "race condition" -- it grabs the CPU and just runs off with all the resources. Usually that's something like going into an infinite loop, with no break point, nothing to tell the thing it has done its job now and can quit running. Opera's wrapper does this pretty often.

The browser also goes into a race condition quite frequently on JavaScript. This would be something I would report through the channels, except it's not consistent. One time I view a site and all is well; next time it goes crazy. Opera has done this since version 5.0, so it's not as if they haven't had time to fix it.

Over all, I'd say Opera (the company) does the same thing a lot of companies on the Net do, not taking Linux and BSD seriously. It's part of why Opera isn't all that popular on Linux and BSD.

Wow. I was beginning to wonder if it would ever be right. I've been having to use Firefox for posting here almost from the beginning because Opera for FreeBSD and SuSE were broken in various ways. For example, the cookies dialog was missing from the beta. I could either accept all cookies or none. Now I can "train" Opera which cookies I want and which to ignore.

However, I won't be rude... Yeah, I will. It's about time, Opera!

Now if I could just get the MyOpera site to work properly....

Update: This note is added from my eMac, also running Opera 9.5. Very nice on the Mac.

This has been an on-going issue since Opera 5: On some particular page with heavy scripting and active media, Opera on Linux will hang, lock up, and enter a race condition. Same page in Opera on any Windows platform doesn't hang, just bogs the system a few moments. I'm beginning to think Opera coders, with all the thousands of bug reports they get with page URLs (a couple dozen from me) still have no clue about Linux. No, it's not even about bad web pages, because other browsers on both OSes do just fine.

It's happened on some of the most diverse hardware, under numerous different distributions and types of Linux, on all versions I've tried so far. I quit filing bug reports. That's because nothing gets fixed. I still love this community and the blog here, but having to go to the commandline to kill Opera should have ended a long time ago. This is partly reflected in the same attitude Opera had about FreeBSD. They were still building Opera for FreeBSD on a machine running FreeBSD 4.3, long after FreeBSD 6.0 was released. Indeed, even on FreeBSD 4.11, stuff that old didn't work. Lots of features in Opera were broken for the longest time.

I'm not sure what the people in charge expect. They seem open to the idea of building Opera for all sorts of oddball operating systems, but want a huge wave of support from users to do it. Then, when technical advice comes along, they refuse to listen. They most certainly won't let folks outside their paid corporate offices see any portion of the code to help them. So if they can't afford to pay someone who really understands the platform, they just tease and offer a broken product.

Yeah, it's free. What did I expect? I'd gladly pay for a browser that worked, and I've bought versions of Opera in the past when that was what it took to get rid of the advertising bar. Nobody's perfect, least of all me, but I get really discouraged when the same problem shows up four full version numbers later.