Menus are fixed and so is startup. Thanks for that! :-)

Scrolling is also improved. Definitely seems faster here.

There's a new bug. Resizing the window to the full width of the screen (1024 pixels) leaves a vertical gray strip of about 100 pixels to the right and a similar, narrower strip at the bottom (below the status bar)

Some other issues, most I believe are known:

* Double Cmd+W closes Firefox, without any prompts. It makes closing more than one tab at a time (with this keybinding) very risky.
* Preferences take a very long time to apply - felt like half a minute (on a Duron 850Mhz)
* Resizing is still slow, how about using frame resize instead of "live" one?
* Maximizing the window doesn't take into account the possibility of a horizontal Deskbar placed at the bottom or at the top of the screen. It makes the window go over/under the Deskbar.
* Import Wizard should be disabled on BeOS. It currently shows Netscape 4.x as an option. If we get Net+ or Opera importing, it might be worthwhile to put this option back, but I don't think either is worth the bother.
* Zooming text (Cmd+"+"/"-") has a delay of a several seconds when first used. It's reasonably responsive afterwords.
* In the Add Bookmark dialog, Create In drop-down doesn't work.

If you want me to file any of these as a new bug at b.m.o, please let me know.

Other than the above, a usable version. Thanks for working on it,

Prog.

(Reply) (Thread)

* Resizing the window to the full width of the screen (1024 pixels) leaves a vertical gray strip...

Could not reproduce that. Screenshot?



* Double Cmd+W
Hmm, must check how it is in Windows.

* Preferences take a very long time to apply
Havn't noticed that, maybe your profile has been updated and has grown big?

* Resizing is still slow, how about using frame resize instead of "live" one?
I'm not very fond of disabling things when they should be improved, and this build is for testing the improvements made to it. It's not perfect.

* Maximizing
It doesn't take deskbar into account whereever it is, personally I think that's ok.

* Import Wizard
Don't see any meaning in doing that, it's already there and someone might want to try importing a windows one. It's just extra work disabling.

* Zooming text
Havn't tried that.

* In the Add Bookmark dialog
You mean selecting another folder than 'bookmarks'?

(Reply) (Parent) (Thread)

> Could not reproduce that. Screenshot?

When I get back home, later today.

> Hmm, must check how it is in Windows.

Pressing Ctrl+W twice only closes two tabs in Windows. It's the same with OS X (using Cmd+W).

> Havn't noticed that, maybe your profile has been updated and has grown big?

I did quite a few changes to the fonts configuration, perhaps it's realted. I'll recheck it.

> It doesn't take deskbar into account whereever it is, personally I think that's ok.

Well, I mapped* Winkey+Z to bottom-screen Deskbar and Winkey+A to upper-right Deskbar and I switch a lot between them (due to the Deskbar's tendency to get in the way). In this configuration, this problem is quite frequent.

* See fourth comment here: http://bebits.com/talkback/3301

* You mean selecting another folder than 'bookmarks'?

Exactly. It should be possible without extending the dialog. That's what this drop-down menu is for.

Prog.

(Reply) (Parent) (Thread)

* You mean selecting another folder than 'bookmarks'?
Didn't know it was broken, must look that up, it may be due to DnD as well. Open your JavaScript console and try to add to see if it gives an error.

(Reply) (Parent) (Thread)

After rebooting, I can't reproduce the gray strips bug. I don't know what triggered it earlier.

Changing fonts takes a long time. Try to change fonts for several languages and you should see probably see it.

Prog.

(Reply) (Parent) (Thread)

Starts, runs under both 5.0.3+BONE and 5.1d0 Dano. No scrolling problems (which gcc and optimizations did you use?)

Regarding bugs blocked by 266252

104729 - no drag'n'drop. No surprise here.
169506 - not sure how to test
239813 - transparency seems to be working (www.tele2.ee menus)
242315 - sorry, can't test: Koki?
276373 - not quite sure how to test; was able to navigate to folders on multiple volumes from the Tools/Options/Download selector by navigating up to Desktop then back down into another volume. Able to navigate to folders on both BFS and FAT32 volumes. Let me know how else to test this, please.
276378 - Bug seems to be resolved, to a point. Tools/Options/Download selecting "Desktop" then "show folder" does nothing, which is proper for my version of Tracker. Selecting any other folder shows the appropriate folder. It might be nice to issue an information message when selecting "Desktop" then "show folder" so people know not all versions of BeOS Tracker will display the desktop in a window. We no longer display the incorrect folder.
277015 - hang on some SSL sites still happens as before.

Concerning Prognathous' comments:
- could not duplicate the "full screen / grey bar" problem.
- cmd-W closes open single window without warning. Wndoz does the same thing. Seems to be intended behavior.
- prefs subjectively no difference on my 1.8G AthlonXP.
- Maximize over horizontal bottom deskbar - yup, that's true but it does it over a top deskbar, too. This behavior is identical to other BeOS apps (incl Net+). Not sure this is a problem w/firefox.
- Import wizard: probably a good idea, but there's always the possibility someone might want to import Netscape prefs off a FAT32 or NTFS volume. I may play with this at some point. It doesn't hurt to have it here, IMO.
- no delay with text zooming on my machine (sorry prog - could be hardware related?)
- bookmarks/bookmark this page dialogue. There's what looks like a drop down with no arrow on the "Create in" window; does nothing, but the arrow next to it works fine.

Long post, but this is a fine build. It's noticeably slower than fyysik's last "-o3" speed-demon mozilla build. But of course, this doesn't crash under Dano. 8^)

tqh, this is a great accomplishment IMO. thank you!

(Reply) (Thread)

Japanese inline patch is not applied due to my changes in window code. It's good to know fyysik's code is fast, because it's other patches, which should be able to be combined.

Good comments btw.

(Reply) (Parent) (Thread)

"Good comments btw."
Thanks! I figure if I can't dev, I have to deliver quality feedback if I'm to help at all.

(Reply) (Parent) (Thread)

> could not duplicate the "full screen / grey bar" problem.

Yeap, it seems to be gone now.

> cmd-W closes open single window without warning. Wndoz does the same thing. Seems to be intended behavior.

Try to press Cmd+W twice in a window full of tabs (each with content) -> The entire window is closed.

> prefs subjectively no difference on my 1.8G AthlonXP.

I tried it with a fresh profile, and it is indeed much better, but still noticeably slower than Windows.

> Maximize over horizontal bottom deskbar - yup, that's true but it does it over a top deskbar, too. This behavior is identical to other BeOS apps (incl Net+). Not sure this is a problem w/firefox.

It's a problem that effects other applications.

> no delay with text zooming on my machine (sorry prog - could be hardware related?)

I'd say that on a 850MHz machine, it should be immediate. It is on Windows.

Prog.

(Reply) (Parent) (Thread)

CMD-W, I'll have to look into it.

The maximize problem is a feature request, I will not have anything to do with it as I don't find it useful.

The other comments are very hard to do anything specific about. It's a bit subjective of your hardware and comparison to windows, all we can do is try to improve speed and I think we can do quite a bit there.

(Reply) (Parent) (Thread)

CMD-W works as expected on my machine. It just closes the tab and then the next one until there are no more.

(Reply) (Parent) (Thread)

I double-checked the CMD-W behavior since my 1st test didn't have content in each window. It works as tqh describes, closing 1 window per command. Here's a thought: do you use the default "ALT" key as your CMD, or have you changed preferences to use "CTRL"? I use the BeOS default "ALT" key. If you're doing something different, it could be related.

(Reply) (Parent) (Thread)

I use the default - Alt as Cmd. I'm quite sure fyysik knows the problem, I recall mentioning this on beshare some time ago.

What happens if you hold the Alt and double-tap the 'W'? What happens if you hold the 'W' for a longer period, say a second?

Prog.

(Reply) (Parent) (Thread)

If I hold down CMD and tap W, behavior is expected. If I hesitate on "W", windows close in very quick succession as you describe. This behavior *is* different than Wndz. As an experiment, I opened StyledEdit and held down "j". A stream of letters appeared very quickly. Under Wndz, there's a hesitation between the 1st occurrance of a letter and the stream to follow. Under BeOS, hesitation is much shorter. This seems more related to the BeOS input server than Bezilla. No doubt, though, it *can* cause headaches if you hesitate on the keys.

(Reply) (Parent) (Thread)

What happens is that I use grouped bookmarks in Mozilla or the Open in Tabs feature of Firefox. In both cases I start with about 20 open tabs. I then read the first tab, hold down Cmd, press W, this tab closes, the next one appears, I read it, press W once more, read the next one and so on. Throughout this time, Cmd (Alt) remains pressed.

The bug happens if I press W twice in succession (to close two tabs). The current window (not the whole browser) then closes, instead of the current tab. It doesn't happen on Windows, Linux, OS X or OS/2, I can tell you that.

Prog.

(Reply) (Parent) (Thread)

This behavior seems directly related to keyboard repeat settings. I played with the keyboard preferences and by increasing the delay before repeat starts, I was able to approximate the behavior of firefox under windows. by decreasing the repeat rate, I could control the speed with which everything closes. I could press W twice in succession with no troubles, especially with the delay before repeat increased from the "short" default. I hope this helps.

(Reply) (Parent) (Thread)

> This behavior seems directly related to keyboard repeat settings. I
> played with the keyboard preferences and by increasing the delay before > repeat starts, I was able to approximate the behavior of firefox under
> windows.

I tested both ends of the keyboard repeat setting in WinXP, and couldn't reproduce this problem. I both cases, each Ctrl+W closed a single tab, not the whole window. I actually can't see how it's related to keyboard repeat rate, since in all cases (including in BeOS), we're dealing with two very distinct and separated presses of the 'W', staccato if you want. The modifier key (Cmd/Ctrl) is pressed throughout, but there's no need to hold the 'W' pressed to see the problem in BeOS.

> by decreasing the repeat rate, I could control the speed with which
> everything closes. I could press W twice in succession with no
> troubles, especially with the delay before repeat increased from the
> "short" default. I hope this helps.

How fast could this repeat be? There's no way it closed a window with dozens of tabs in less than one second? If it does, like in my BeOS case, it sounds like a bug.

Could this be relate to SpicyKeys? Do you use it? I always do (can't do without it).

Prog.

(Reply) (Parent) (Thread)

I don't use SpicyKeys - never have - so it's possible this is related. Based on your experiences, I'm not certain I want to install and test. 8^)

(Reply) (Parent) (Thread)

Well, it's not SpicyKeys. I uninstalled it and the problem is still there :-(

BTW, you really should install SpicyKeys, it's an absolutely indispensable utility, especially when coupled with other apps.

Prog.

(Reply) (Parent) (Thread)

Hi,
I tried this build the other say and after double clicking on the firefox file to run it, Firefox crashed before completely loading. I'm using BeOS R5 + BONE. The system I tested it on is an AMD K6-2 500. I'm starting to think there's something wrong with my BeOS installation since many programs that are said to work with BeOS R5+BONE are not working or just crashing. Not sure what the problem might be. I get this from the debugger after it crashed:

loading symbols
invalid opcode occurred
nsDependentCSubstring::Rebind(nsACString const &, unsigned int, unsigned int):
Rebind__21nsDependentCSubstringRC10nsACStringUiUi:
+0035 edf43099: * f8470f cmova %eax, %edi
firefox-bin:

Is it possible that my Firefox Profile folder is corrupted, causing it to crash at the beginning. I tried installing a new emptyprofile, but all the ones I try to download like the one off Bebits don't unzip properly saying the file is corrupt. I've downloaded different versions from other people, but end up with the same problem.

(Reply) (Thread)

Hmm, I'm not sure if 'i686'-archticture I used when compiling supports K6 or K6-2... No it doesn't. So this build isn't working with your machine. I'll do the next build with i586. Apparently K6-2 doesn't have conditional move 'cmova' and that is what you are seeing. I had not thought of this before building.


Btw, if you have problems with several other apps you could always test your memory: http://www.memtest86.com/

(Reply) (Parent) (Thread)

Thanks for the info tqh. Also, if you build the next version with i586, will that effect the browsers performance on other systems that use i686 or higher prcessors?

(Reply) (Parent) (Thread)

I havn't noticed much difference, and I think we should optimize for the lowend machines, not the higend ones anyway.

(Reply) (Parent) (Thread)

"I think we should optimize for the lowend machines"
This is always the dilemma when developing - since the BeOS community is small to start with, we always have to aim for the "lowest common denominator" in order for our efforts to benefit a significant number of people. Ideally, IMO, we wouldn't even have BONE and Netserver versions; just one fast Netserver version that runs cleanly under BONE, Dano and Zeta.

But you've now given me one more testing variable - I have an old K6-2 machine hanging around; maybe I'll load BeOS and give new builds at least a cursory test... "in my spare time." ;)

(Reply) (Parent) (Thread)

there were investigations about optimization flags
actually, better/faster code is generated with gcc if cpu is set to even 386, and only certain optimization flags to 586 or 686

(Reply) (Parent) (Thread)

Yes but if you've seen all the awful things that is done to get netserver to even work, you'd never use that one.

(Reply) (Parent) (Thread)

K6 is known to be problematic. See https://bugzilla.mozilla.org/show_bug.cgi?id=233815#c3 (Linux)
and https://bugzilla.mozilla.org/show_bug.cgi?id=225472 (OS/2)

Prog.

(Reply) (Parent) (Thread)

Second link should have been https://bugzilla.mozilla.org/show_bug.cgi?id=239947

Prog.

(Reply) (Parent) (Thread)

CougarGT, I have a K6-2/533 and an old Asus P/E-P55t2P4D dual-CPU AT-style mainboard - part of a project I'll never complete. If you've want to experiment with a dual-CPU BeOS system, you've already got one K6-2 so you're most of the way there. Contact me directly by removing the obvious if you're interested: (nospam)dshelton(att)san(dot)rr(dot)com. Sorry for the OT post.

(Reply) (Parent) (Thread)

was this done with 2.9 or the newer 2.95.3?

(Reply) (Thread)

I only use Olivier's gcc (2.95.3) as it seems to be far better than the rest. His gcc is very very good.

(Reply) (Parent) (Thread)

Interesting: this version runs fine under 5.1d0 Dano. Using the same compiler, fyssik's 1.7a -O3 build crashed with scrolling. So maybe it's the optimization level and not the GCC version, that caused the scrolling crash.

(Reply) (Parent) (Thread)

this problem was already explained.
Even here.
Workaround and fix are known

(Reply) (Parent) (Thread)

(Reply) (Parent) (Thread)

Yes, and I *knew* this. I apologize for wasting time with this thread. Thanks for making my brain function.

(Reply) (Parent) (Thread)

Hi guys,

Nice build! It feels really snappy; as a matter of fact, this is clearly the fastest Firefox I have ever used so far. Using it on Zeta on and AMD 2.2GHz with Neo + SP1, and no problems so far.

Regarding the Japanese inline input patch, will this need to be revised?

Keep up the good work!

Koki

PS: The ttps://login.personal.wamu.com still does not work for me though.

(Reply) (Thread)

I don't think the patch needs changes, I should have applied it before doing my own changes to nsWindow.

(Reply) (Parent) (Thread)