I have uploaded a new unofficial version KM-16-F3620 ,
based on the updated Firefox 3.6.20 , Gecko 1.9.2.20 , date 20110803 .
Firefox 3.6.20 is supposed to be the last updated Firefox of the 3.6 series. As K-Meleon is not updateable using newer Firefox series, this may be the last possible update using the k-meleon.exes that are available at the moment.
It is nevertheless possible that Mozilla will release further updates
for Firefox 3.6, as many business customers will want to continue using the older Firefox versions.
Note : This version needs the 80's runtimes. Download MSvce8dist.exe from
http://jujuland.pagesperso-orange.fr and install it.
As kplugins from K-Meleon 1.5 are being used, you will also need
the files mfc71u.dll, msvcp71.dll and msvcr71.dll in your system or in your K-Meleon folder.
If you need these files, look for them in the internet.
The profile will be created inside the K-Meleon folder.
If in Linux you want it inside the ".wine" folder, because of
user rights problems, rename the file profile.ini to
profile.ini-renamed before the first start.

Download from here :

http://kmext.sourceforge.net/fred/KM-16-F3620.zip

Regards to all

Fred

Firefox 3.6.21 will be on the road at september, 27.

http://blog.ffextensionguru.com/2011/08/16/firefox-3-6-20-released/

o.O

that's good news for those using FF 3.6, margarita. FF7 final & FF8 beta will also be released on Sept. 27 along with version 3.6.21.

@Fred: I'm wondering if you can make an unoffcial KM-16-S2015pre build with the latest gecko 1.9.1.20pre build.



Edited 1 time(s). Last edit at 08/19/2011 03:57AM by 4td8s.

@ 4td8s

OK. I have uploaded a last 1.9.1.20pre version KM-16-S2015pre.
Please be aware that this version may already have security
vulnerabilities and cannot be updated anymore because
Gecko 1.9.1.x has reached its end of life for Seamonkey
and Firefox.

Download here :

http://kmext.sourceforge.net/fred/KM-16-S2015pre.zip

Regards

Fred

Quote
Fred

Gecko 1.9.1.x has reached its end of life for Seamonkey
and Firefox.

True but perhaps note that mozilla-1.9.1 Branch development on Firefox 3.5 and Gecko 1.9.1. was last changed at Tue Jul 05 14:29:34 2011 -0700.

Source for info at: http://hg.mozilla.org/releases/?sort=lastchange

p.s. Currently there is no newer GRE source code for a fully functional K-Meleon available. So this GRE based on the Source tarball available from the linked page is the newest possible.

Quote
Fred
I have uploaded a new unofficial version KM-16-F3620

Thanks for you great work. This is the first of your unofficial builds that I've tried and it's been pretty much perfect. I may be imagining things but this just feels snappier than the 1.6b2 that I've been running.

In normal use I've found only a couple of things you may want to know about.

1. Login dialog boxes don't work. When a site (or in my case a router) tries to display a Windows style dialog box for login the box doesn't appear. Instead I go straight though to the target site and, of course, get rejected as not authenticated.

I checked KM's error console and found errors for nsLoginManagerPrompter.js. Removing this file from the Components folder seems to allow login boxes to work normally again.

2. Spacebar doesn't scroll on the first page opened in a new tab. Mouse/trackpad scrolling is okay but the spacebar won't scroll until a link is clicked on the page. Then it works fine in that tab. This probably isn't a big deal for most people but I'm in the habit of opening everything in tabs and so used to scrolling down that way I'm always tripping over it.

I'm using Windows7 32bit if that makes any difference.

Once again thanks for your effort.

Yes, Fred's 1.6 builds are snappier than 1.6b2. And they usually don't have the annoying random moouse down selection bug that 1.6b2 has, at least under W98 KX.

pretty stable too on xp2. love the speed............................

I have tested it under Linux, and all the little defaults under wine seems to be the same as before.

Like 1.6.0b2, no problem to launch itfrom a shortcut (which isn't the same with 1.7.0a)

I have found a reason of the messages when launching in console:

err:toolbar:TOOLBAR_GetImageListForDrawing bitmap for ID 0, index 0 is not valid, number of bitmaps in imagelist: 0

It occurs only if menu is displayed as a toolbar.

Some others warnings in console, probably to to missing or outdated dll :

err:ntdll:NtQueryInformationToken Unhandled Token Information class 15!
fixme:file:MoveFileWithProgressW MOVEFILE_WRITE_THROUGH unimplemented
fixme:advapi:SetNamedSecurityInfoW L"E:\\Program Files\\KM-16-S2015pre\\components\\xpti.dat" 1 536870916 (nil) (nil) 0x134cd4 (nil)
fixme:iphlpapi:NotifyAddrChange (Handle 0x111e9b0, overlapped 0x111e998): stub
fixme:file:MoveFileWithProgressW MOVEFILE_WRITE_THROUGH unimplemented
fixme:advapi:SetNamedSecurityInfoW L"E:\\Program Files\\KM-16-S2015pre\\Profiles\\zt19itwu.default\\prefs.js" 1 536870916 (nil) (nil) 0x17d47c (nil)
fixme:file:MoveFileWithProgressW MOVEFILE_WRITE_THROUGH unimplemented
fixme:advapi:SetNamedSecurityInfoW L"E:\\Program Files\\KM-16-S2015pre\\Profiles\\zt19itwu.default\\prefs.js" 1 536870916 (nil) (nil) 0x190164 (nil)
fixmeystem:SetProcessDPIAware stub!
fixme:ntdll:NtLockFile I/O completion on lock not implemented yet
fixme:keyboard:RegisterHotKey (0x20080,1,0x00000001,C0): stub
fixme:resource:GetGuiResources (0xffffffff,0): stub
fixme:file:MoveFileWithProgressW MOVEFILE_WRITE_THROUGH unimplemented
fixme:advapi:SetNamedSecurityInfoW L"E:\\Program Files\\KM-16-S2015pre\\Profiles\\zt19itwu.default\\prefs.js" 1 536870916 (nil) (nil) 0x1de064 (nil)

1.6.0-S2015 seems to be faster than 1.6.0b2, probably due to new used gecko ?

No more need the button URL, as there is no more crash with URL address toolbar.

A+

---

Mozilla/5.0 (x11; U; Linux x86_64; en-US; rv:1.9.1.17pre) Gecko/20101211 Ubuntu/12.04 (precise) K-Meleon/1.6.0
Mozilla/5.0 (x11; U; Linux x86_64; fr-FR; rv:1.9.2.14pre) Gecko/20101219 Ubuntu/12.04 (precise) K-Meleon/1.7.0

---

Web: K-Meleon Extension Setup (French) and (English)
Web: http://jujuland.pagesperso-orange.fr/
Mail : alain [dot] aupeix [at] wanadoo [dot] fr

---

Edited 1 time(s). Last edit at 08/26/2011 04:09AM by JujuLand.

Quote
JujuLand
1.6.0-S2015 seems to be faster than 1.6.0b2, probably due to new used gecko ?

The newer code base is not the reason.

The GRE is compiled to favor speed (-O2) not to favor small size (-O1).

The later is often used in official K-Meleon GREs.



Edited 1 time(s). Last edit at 08/28/2011 08:53PM by guenter.

I think there is more to it than that, I do notice the change in js speed on some sites, i believe there was a big jump in js performance between these versions...

but perhaps more significantly, many of Fred's versions don't have the random mouse select bug, this bug also seems to include random delays where nothing appears to be happening. And I have feeling it is due to some compilition switch or library change between the two versions.

F3618 is one that has the bug while S2014 is the only one I am absolutely sure does not have the bug. S2014 is rock stable and responsive on my machine under W98KX... And I immediately noticed improvement from 1.6b2 when I installed it, and I'm still using now!

Quote
snuz2
I think there is more to it than that, I do notice the change in js speed on some sites, i believe there was a big jump in js performance between these versions...

Favoring speed over size gives a boost of up to 20 odd %.
Most users need about 25% difference to notice.
That is no marginal improvement but not a big one either.

Just to make sure we mean the same thing.
That is for the same GRE major version.

But sub versions of GRE 1.9.2 or faster than sub versions of GRE 1.9.1 and so on.


p.s. I habitually run new GREs over an older V8 JavaScript Benchmark Suite (version 5 or older) to see whether they are sound. That is my score.

AFAIK mozilla.org retrofits no JavaScript engine speed improvements into their older GRE versions. They only fix security problems.



Edited 2 time(s). Last edit at 09/02/2011 10:38PM by guenter.

yeah, they don't retrofit that unless it comes by accident. but there were 2 big jumps in speed in FF, one was v4 and one was earlier, and in between they kept ramping up the speed in the intermediate versions. I was thinking of the js engine in the Fred's 1.7 hybrid, which is almost like FF4. but that may not be in either of the builds I compared, i don't remember.

My main point is that builds that don't have the random mouse selection bug also run faster. whether that is because they were built for speed rather than size, or it is just a side effect of different compilation options or library versions, I don't know.
But a bug that
- exists in 1.6B2, but not one of Fred's builds, and then reappears in his later builds
- has to do with handling input/output, and seems to generally slow the execution
doesn't seem like an error in KM code itself. It seems like a build configuration problem. i don't think it is a direct result of compiling for speed/ size itself but maybe a library changes when that or some other option is selected and something doesn't work right anymore. could just be my imagination.

maybe selecting compact size activates some memory scavenging mechanism that causes input to be dropped and random pauses, but i doubt it.

anyway, builds that don't have the random mouse select bug seem to run much better in general on my machine and i think that that bug is a symptom of a larger performance killing problem. builds without the bug also seem to respond to user input faster and more consistently. that's the point i'm trying to make.



Edited 1 time(s). Last edit at 09/03/2011 12:23PM by snuz2.

Maybe B2 some slower then builds from Fred, because 1.6 b2 compile with debuging support.

Quote
margarita
Firefox 3.6.21 will be on the road at september, 27.
http://blog.ffextensionguru.com/2011/08/16/firefox-3-6-20-released/
o.O

Wow, it's Firefox 3.6.23 now!
https://wiki.mozilla.org/Releases/Firefox_3.6.23
At least one more time with KM-16-F362x... probably the best unofficial branch ever.
IMHO, of course.
o.O

Margarita,

You can thank Asa's big mouth for the continued existence of the Gecko 1.9.2 line and yes a KM-16-F3623 would be very much appreciated IMO.

<g>

N

Firefox 9.0 with Type Inference is much, much faster than anything based on Firefox 3.6's outdated engine, Gecko 1.9.2.

http://www.conceivablytech.com/9229/products/a-year-in-browsers-dazzling-performance-gains/2

http://www.conceivablytech.com/wp-content/uploads/2011/09/JAVASCRIPT.jpg

http://www.conceivablytech.com/wp-content/uploads/2011/09/HTML5.jpg

524% faster, no matter what compiler optimization tricks you use, you'll never be able to compete with Firefox 9.0 in performance.

Quote
FX9
Firefox 9.0 with Type Inference is much, much faster than anything based on Firefox 3.6's outdated engine, Gecko 1.9.2.

http://www.conceivablytech.com/9229/products/a-year-in-browsers-dazzling-performance-gains/2

http://www.conceivablytech.com/wp-content/uploads/2011/09/JAVASCRIPT.jpg

http://www.conceivablytech.com/wp-content/uploads/2011/09/HTML5.jpg

524% faster, no matter what compiler optimization tricks you use, you'll never be able to compete with Firefox 9.0 in performance.

Give me 80 million a year in income and I too could come up with new releases. Personally I like KM for KKO's delightful macro language. As for FF9, I have enough trouble with FF 7.05beta and my broken extensions there to even think of going on to an alpha version.

N

Quote
FX9
Firefox 9.0 with Type Inference is much, much faster than anything based on Firefox 3.6's outdated engine, Gecko 1.9.2.

http://www.conceivablytech.com/9229/products/a-year-in-browsers-dazzling-performance-gains/2

http://www.conceivablytech.com/wp-content/uploads/2011/09/JAVASCRIPT.jpg

http://www.conceivablytech.com/wp-content/uploads/2011/09/HTML5.jpg

524% faster, no matter what compiler optimization tricks you use, you'll never be able to compete with Firefox 9.0 in performance.

There is no Firefox 9.0 yet.

Nightly is at Version: Mozilla/5.0 (Windows NT 5.1; rv:9.0a1) Gecko/20110911 Firefox/9.0a1

And Google Chrome Development version is much faster than Firefox Nightly


& both have a clumsy handling and interface compared to an old & slow K-Meleon.



Edited 2 time(s). Last edit at 09/12/2011 05:04AM by guenter.