掲示板 過去ログ
(2011年7月1日 〜 2011年12月31日)

In portable mode, my builds use LocalAppData folder in place of %LocalAppData% folder, so LocalAppData may have the necessary data to run my build.

Yes, I know this. But Earlybird shouldn't crash in any case. It should create new LocalAppData folder.
And now I'm pretty sure this crashes related to the new tmemutil.dll. Tried Earlybird 9 with tmemutil.dll from Earlybird 8 and all work just fine. No crashes.

Thank you for all the information. I will update tmemutil.dll later.

Firefoxの既定の動作では、複数のFirefoxを同時に実行できません。メインウィンドウが複数作成されて同時実行できたかのように見える場合でも、それらは全て最初に起動したfirefox.exeのプロセスです。よって、最初に実行したfirefox.exeのプロファイルがそのまま使われ続けます。

複数のfirefox.exeを同時に起動できるようにするには、それぞれの起動オプションに-no-remoteを付けてみていただければと思います。
firefox.exe -no-remote

ポータブルオプションを付けた場合に作成されるフォルダの中身ですが、「-profile""」を指定したときと同様のフォルダ構成（プロファイルが1フォルダに収まる）にすることは可能でしょうか？

ポータブル化した各ビルドのプロファイルを、一つのフォルダ内に入れたいということでしょうか?

ポータブル化した各ビルドのtmemutil.iniのPortableDataPathに同じ場所を指定すれば、一つのフォルダにまとめることが可能かと思います。
http://www1.plala.or.jp/tete009/software.html#TMOZFAQ_INI

こんにちは。
横から失礼します。

tmemutil.iniによるポータブルオプションを利用した場合のプロファイルについて，ちらっと試してみた限り「-profile "プロファイルフォルダー"」オプションで起動した場合と同様に，プロファイルマネージャーに登録される形ではプロファイルを利用されないように見受けられます。
また，その状態で別々のフォルダーに存在する異なるバージョン（9.0，9.0.1）のfirefox.exeを起動しても，一瞬だけ9.0.1のプロセスが現れますが，実際に開いたウインドウでFirefoxについてからバージョンを確認すると9.0と表示されています。

Firefox（やThunderbird，SeaMonkey）で異なるプロファイル，異なるバージョン（というか異なるexe）を同時起動するためには，プロファイルを分けるだけでは不十分で，「-no-remote」オプションを用いて起動する必要があるように思われます。
別々のフォルダーのfirefox.exeに「-no-remote」オプションをつけて起動を試されてはいかがでしょうか？

P.S.テテさんの見解もいただきたいところです

あっ！書き込んでいるうちにテテさんの書き込みが(^_^;)

匿名さんが期待しているのは，「-profile "プロファイルフォルダー"」オプションを用いた場合と同様に，すべてのプロファイルが単一のフォルダー下に格納される形ではないでしょうか？

tmemutil.iniによるポータブルオプションの場合には，指定のフォルダー下に“AppData”“LocalAppData”フォルダーが作成され，その中にさらに深い階層でデータが格納されているようです。

そういうことでしたら、公式ビルドと同じように可能です。

(PortableDataPath)\AppData\Mozilla\Firefox\profiles.iniを開いて、例えば以下のようにしてプロファイルの場所を指定することもできます。

[Profile0]
Name=default
IsRelative=1
Path=../../../xxxxxxxx.default

すんません、勘違いしてました。(^^;
もちろん以下のようなこともできます。
firefox.exe -no-remote -profile "(プロファイルのパス)"

http://bit.ly/uwOE7I に書いた方法のまとめ。

PortableDataPath直下のHogeHogeというプロファイルフォルダを使いたい場合、以下のようにします。

1. (PortableDataPath)\AppData\Mozilla\Firefox\profiles.ini を開く。
2. プロファイルのパスを、以下のように相対パスで指定する。
[Profile0]
Name=default
IsRelative=1
Path=../../../HogeHoge

なお、PortableDataPath直下のAppDataとLocalAppDataは、それぞれ%AppData%と%LocalAppData%フォルダの代わりですので、プロファイル以外のデータも保存されている大事なフォルダです。

テテさん、アイヴァーンさん、回答ありがとうございます。
色々言葉足らずで、申し訳ありませんでした。

やはり-no-remote を付けなければいけないのですね。
ポータブルという言葉から、複数バージョンの使い分けも期待してしまっていました。

tmemutil.iniがせっかくあるので、そこで -no-remote の設定も出来ると有難いのですが、
そのようなことはお考えでは無いでしょうか？

アイヴァーンさん

「-profile "プロファイルフォルダー"」オプションで起動した場合と同様に，プロファイルマネージャーに登録される形ではプロファイルを利用されないように見受けられます。

ここでお聞きすることでは無いかもしれませんが、プロファイルマネージャーを使って作成した方がよろしいでしょうか？
プロファイルマネージャーに登録されない欠点などあれば、ご教授いただければと思います。
他のビルドではあまりよくわからず、プロファイルを作っていたので何が違うのか分かっていません。

テテさん

なお、PortableDataPath直下のAppDataとLocalAppDataは、それぞれ%AppData%と%LocalAppData%フォルダの代わりですので、プロファイル以外のデータも保存されている大事なフォルダです。

そうだったのですね。
\AppData\Mozilla\Extenshons\
には何もデータが無く
\AppData\Mozilla\Firefox\profiles\hogehoge\extensions
の方にadd-onのデータは入っているようですし、
\LocalAppData\Mozilla\Firefox\profiles\hogehoge\
とprofilesが別れていて、他のデータは無いように見えたので
なぜ階層だけ深いのかが疑問でした。

自分でも検索して調べたいと思います。
お答えいただきありがとうございました。

匿名さん，こんにちは。

“ここでお聞きすることでは無いかもしれませんが、プロファイルマネージャーを使って作成した方がよろしいでしょうか？”
プロファイルマネージャーに登録される場合には，必然的にそのシステムで固定的に使うこと（＝通常の利用形態）になるかと思います。
ポータブル化して利用する場合にはプロファイルマネージャーを利用する必要性がなく，逆にプロファイルマネージャーによるプロファイル管理はできなくなるかと思います。

Ivarn（アイヴァーン）さん

必然的にそのシステムで固定的に使うこと（＝通常の利用形態）になるかと思います。ポータブル化して利用する場合にはプロファイルマネージャーを利用する必要性がなく，逆にプロファイルマネージャーによるプロファイル管理はできなくなるかと思います。

変な仕様だと思いますが、そうなのですね。ありがとうございました。

tmemutil.dllベータ版で、tmemutil.iniの[Env]セクションに環境変数を指定できるようにしてみましした。
http://tete009.seesaa.net/article/245010301.html#A245010301

例えば、[Env]セクションにMOZ_NO_REMOTE=1を指定すれば、起動オプションの-no-remoteと同じ動作になると思います。

[General]
Portable=0
PortableDataPath=PortableData
CreateCrashDump=0
GdiBatchLimit=0
ProcessAffinityMask=0

[Env]
MOZ_NO_REMOTE=1

XRE_PROFILE_PATHでプロファイルフォルダのパスを指定できるようです。

XRE_PROFILE_NAMEでプロファイル名を指定できるかと思って試してみましたが、成功しませんでした。

テテさん、希望を採用して下さりありがとうございました。
さっそく使わせて頂きました、今のところ特に不具合は無いようです。

ご報告くださりありがとうございました。
今、Firefoxで使用されている環境変数について調べていますが、開発者向けに用意されたものが多く、ユーザー向けのものは少なそうな感じがしています。

With Mozillas first 64bit RC? o_O

x64 more slow with x86 version in case firefox. x64 is more faster in emulator in case dolphin x64 run very more faster then x86 version

At this point I have no plans to build 64-bit versions.

To my regret, Firefox optimized by the 64-bit compiler of VS2005 is unable even to run.

I would like to build FF using My prebuilt (MinGW/MinGW64) GCC in future, but no schedule now.

Just heard that Mozilla is expected to start using MSVC2010 for building Firefox starting with Firefox 12. What does this mean for all of us. It's the start of possibly getting an official Firefox 64bit. Their blog is reporting everything is ready to go:

MSVC 2010 Migration (targeting Gecko 12) - khuey

- Builders are ready to go.
- Tests are green.
- Perf numbers look good.
- Client side updater requirements are done.

Uhmm... Why you mean 64bit Fx is useful?

@Pik

Speed! Add the optimization and you have a very fast web browser. It'll be the fastest browser you can imagine.

There's already Flash 64bit, PDF 64bit, Silverlight 64bit, Java 64bit plugins. On a 64bit operating system, everything won't be limited and slowed down by a RAM limit like on a 32bit operating system. Also everything will be made to run more efficiently and faster because the plugins, web browser and operating system will ALL be 64bit.

If there was no difference, then they wouldn't make software and operating systems in 64bit.

Lol. Yes? By the way, I have asking for useful 64bit browser and not for sense of 64bit OS o_O

Examples? IE9 32bit and 64bit comparsions are not examples. Its Ms tricky part of hanky-panky to forced 64bit Windows.

Ok, nothing against 64bit OS, but for a browser its ridiculous. On very fat 64bit Windows (you have >=16GB and more?) EACH 32bit process can becomes FULL 4GB of RAM.

Contrary to my Photoshop :-D, when your browser, your plugins or your PDF-reader consumpting each more than 4GB RAM -> They are broken. But I agree with you, with 64bit soft and much more RAM you cant see, than maybe you 64bit browser and 64bit Flash bred scary mem leakages. All is not only faster, but also more stabler

Fx12 is probably also 64bit. May 2012.Till then, please hit the sack and dont make folks tense.

Hmm...
"Re-reading rules.mk, I was fairly certain we weren't supposed to embed manifests in DLLs unless we explicitly requested them?"

Good to know, Tetsuro. Thanks.

Dr. Tetsuro Kato, Akita University ? Arent you ?

I thought, curiosity is a womens /ailment/...

He is a person with an identical name.

Dr. Tetsuro Kato, Akita University ? Arent you ?

http://jp.linkedin.com/pub/tetsuro-kato/38/640/601

あなたですか？

He is also a person with an identical name.

How about Atelier ?

Relax, dudes and dudetes. I think he's real name is Bill G. He just have too much free time so he's studying japanese and making sweet firefox builds for all good guys. For me for instance.

Yep. From Top3 builder he is the number one. Atta man that make good high-skilled job.

My Earlybird 9.0.1 has the Test Pilot for Thunderbird 1.3.4.

If it has not installed for some reason, you can find it in the directory of distribution\extensions.

You misunderstand me. I meant maybe you added the TestPilot addon by mistake. Because you never included it in your builds IIRC.

I just built Earlybird 9.0.1 in my usual way, so my build has included the Test Pilot automatically, such as an official Thunderbird 9.0.1.

I'm just thinking that all this feedback stuff should be disabled in custom patched builds. 'Coz it can mess up all Mozilla statistics about performance, crashes and so on. It's my very humble opinion of course.

I absolutely agree with you.
I'd like to look for a way to remove the Test Pilot addon from my build.

And maybe you should get rid of the "Submit performance data" option from Firefox as well? I don't know if it's really works or not.

You can remove distribution\extensions

Thank you for the information.
Maybe I can also delete it by removing the definition of SHIP_FEEDBACK.

waterfox and palemoon 64bit is slightly faster than official alpha edition (only 1%-2%).
I think the most fastest sunspider 64bit edition is htguard edition (http://htguard.island.ac/)

The other day I build FF9 64bit edition using tete009 patches (except tmemutil). You can see http://code.google.com/p/pcxfirefox/downloads/list

My 64bit edition has two edition: pgo and betterpgo. In some tests, my pgo edition is faster. In another tests, the betterpgo edition is faster. (betterpgo's DOM test may be faster and real workload may be better, but I don't test it much time.)

If you have not special needs, you can choose htguard edition.

The hype about palemoon (or waterfox) is for me unintelligible. Of all builds, only PcX have similar speed to tete's builds. All the others (for WinPC) are slower. Partly significant slower, partly not better than the original.

And there tete at the moment dont make 64bit versions, only PcX 64bit is noteworthy.

先ほどためしたら、Windows版のFirefox 9.0でも更新機能で9.0.1にアップデートされました。

昨日、該当箇所のソースコードを見てみました。Windows版でどのくらい影響があるのか、自分には分かりませんでした(汗)。でもWindows版でも実行されるコードなので、更新されるだろうとは思ってました。

JS-benches say... NOTHING. "Perceived" tete is for example on http://www.theverge.com/ twice as fast then ayakawa.

Also on the german http://www.spiegel.de/wissenschaft/natur/ tete "load" the pictures much faster than antoher builds.

Only PcX is similar in speed, but it will NOT worsk with SpeedDial (!) With hundrets messages (css) in the error console.

This build have also tons of stuff in /chrome and /components than tete have not (?!)

p.s.:
I have tested PcX's "better-PGO" build. A little Wow, tete Bloody fast, but like I said not Spedddial compatible and with "peculiarly" /chrome and /components.

But clearly, very fast.

Sorry, I don't use SpeedDial. My edition (PcX) 's patches are based on tete009 except tmemutil. My x64 edition is also based on tete009, too.
BetterPGO contains many popular tests and real world test, but I don't know this influence. Other people said that my edition works more fluently, but I also use tete009 edition very fluently.
It seems to me that tete009 is the stablest edition.:-)

"I don't use Speeddial"? :-) It's no biggie to make builds which are /inkompatible/ with this, that and the other Addons. The big exercise is make faster build than Mozilla, but with 1:1 compatibility. Id est, to make *full* build

@tete
It appears to me, Better-PGO are not a bad set.

I have collected the profiling data for PGO in the following way:
http://tete009.up.seesaa.net/board_log/board_12.html#T1279786376.93

Under specific conditions, my builds use multithreaded routines in some functions. For example:

CSS3 text-shadow and box-shadow:
http://lea.verou.me/css3patterns/

Canvas getImageData and putImageData:
https://bug519400.bugzilla.mozilla.org/attachment.cgi?id=426752

Some SVG filter effects:
http://tete009.up.seesaa.net/html/svg_test.html - my site's logo
http://svg.kvalitne.cz/bump/bump.xhtml
http://srufaculty.sru.edu/david.dailey/svg/Turbulence1.svg
http://srufaculty.sru.edu/david.dailey/svg/water3.svg
http://srufaculty.sru.edu/david.dailey/svg/scenes/popcorn11.svg
http://srufaculty.sru.edu/david.dailey/svg/matrixFilter.svg

There are some other changes in my builds, but I don't go into details here.

@PiK

Yea. Because I don't use SpeedDial, I haven't find the incompatibility and I won't fix the bug what I didn't encounter.

About BetterPGO, this is only the combination property improving's attempt. As VC Compiler PGO has many bugs, I don't know whether the BetterPGO edition continuously works all right.

tete009 edition is the only one what has multithreaded canvas/pixman/svg.

@TeTe

btw, Your patches have some incompatible things on VC2010.

jsstr.h : void __movsw is also declared in intrin.h, so you may not declare it again.
jslock.cpp : _InterlockedCompareExchange likewise

on Win64 build, the asm of gfx\2d\DrawTargetD2D.cpp can't be used on VC2010 (some functions change and miss), so I revert it to official.

@xunxun
Is ok. I would only help you, then with this thinking you can NOT go to "forever memory" ;-)

@tete
I have noted your multithreaded routines. I mean only comparsion of your PGO to "better-pgo".
I mean this is /in front/ a first time a PGO than is not badder than your and thus _maybe_ be a value to deign a look. No more

@xunxun
About jsstr.h. On VS2005, when including intrin.h in jsstr.h, a compiler error occurs, so I am doing it.

@xunxun
Lol. "Don't use SpeedDial"? You must not, but a Firefox without 100% compatibility to original is purposeless. Or easier: the build is simply BROKEN

I've no idea. I installed the extension (SpeedDial) just now, but it seems to work well for me. Can you describe it where it is abnormal?

Thanks for testing! All right now. My fault. I am back to tete, but you are not badder ;)

only revert one patch

@xunxun
Which one? Url, man ;)

You can see

http://hg.mozilla.org/releases/mozilla-release/rev/b78fe362789b

Grrr.... :-\
http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/9.0.1-candidates/build1/

Please update , tete !

Please boy, this is nerve-racking... :-\

Thank you for waiting.
I have just uploaded my build of Aurora 9.0.1.

Of course I mean "FF 9.0.1" user and not you (tete)

I understand, you will make also your job or you will sleep or anything you like

Thanks , tete !

@Pik

Are you sure FF3.6.x and thunderbird 3.1.x is the final edition? If so, I want to build them.

Why? ;)

Building it for forever memory.

THANKS!

http://blog.ffextensionguru.com/2011/12/21/firefox-3-6-25-released/

He said 3.6.x is not the end edition, and there will be 3.6.26

final

I will upload my Namoroka 3.6.25 after a few hours.

現在アップされているAurora 9.0 candidates build1 私的ビルドには、WindowsにVisual C++ランタイム8.0.50727.6195がインストールされていないと、components\browsercomps.dllのロードに失敗するという問題があります。

私的ビルドのルートフォルダにある以下のファイルを、componentsフォルダの中にコピーすれば、とりあえず動くようです。
Microsoft.VC80.CRT.manifest
msvcr80.dll
msvcp80.dll
msvcm80.dll

公式ビルドでも、Visual C++ 2005 SP1 再頒布可能パッケージ (x86) (バージョン8.0.50727.762 ?)がWindowsにインストールされていない環境では同じ不具合が発生するため、Bug 713167にて報告しました。エセ英語で。^^;

I used resources editor to edit the manifest
with the version string "8.0.50727.762"
and it works fine now.

Appreciable?
http://stackoverflow.com/questions/1582844/vcredist-x86-dll-and-version-8-0-50727-4053

there wasn't manifest in browsercomps.dll in the previous builds. therefore all was fine.

so you can manually delete manifest form the file or use this http://www.microsoft.com/download/en/details.aspx?id=26347 .

Firefox 8.0 or less versions use custom C-Runtime DLLs such as mozcrt19.dll, so manifest definition for C-Runtime DLL is unnecessary.

But Firefox 9.0 or more versions use C-Runtime DLLs built by MS, so the manifest definition is necessary.

yes. you're right of course. but browsercomps.dll without a manifest still works for me. at least on WinXP SP3.

tico, the "edit manifest and it works" bothers steady... This was not the topic.

Ohh.. I'm so sorry for hijacking your personal thread.

@tete
For your 9.0. Perfomance of partly-cached sites is amazing.

Empty-cache perfomance... Without wows I am testing it yet in comparsion of your 8.0.1.
At the moment I mean 8.0.1 is little bit faster in site "composition"...

After "many" history clears, the "allround" perfomrmance of 9.0 is noticeable better. sql reorg? not a hunch, but ok :)

But I cant see it in 480p without plugin-container 10x crashed...

http://www.cartoonland.de/archiv/best-fails-2011/

Yep. Plugin-container EXTREMELY instable.

In my Aurora 9.0, I removed the resource of type RT_MANIFEST from the browsercomps.dll. This may enable my build to load a browsercomps.dll even if a C-runtime library 8.0.50727.6195 has not been installed in the system.

There are several parts where illegal codes are generated by VS2005's PGO, so about these files, I have disabled PGO by adding -GL- option.

I could see the Flash movie with 480p, enabling plugin-container. My Flash version is 11.1.102.55.

But I cant see it in 480p without plugin-container 10x crashed...
http://www.cartoonland.de/archiv/best-fails-2011/

What your memory consumption of plugin-continaer? It blowed the ram herte

But I have it not compare to original Fx9.0.

p.s.:
Are PGO problems with 9.0 "rather" a VC2005 troubles? I mean, orignal Fx9.0 is also PGO-ized (?) Or not?

Lol... For testing I have comparing your with Ayakawas "hightech" VC10sp1 build.
NO WAY. Comparing to yours on my 3.8Ghz Penryn, its simply slow I think its his PGO/VC on (for??) his CPU (?)

I hope this not ending similar, when you make anytime a VC upgrade :-\ But VC2008sp1 were a option... (?)

In official Firefox 9.0, PGO has not been applied to js module (Bug 605033).

Another "no way". I have must deactivate plugin-container through Youtube&Co. Fly away of Fx every umpteen seconds.

Without container its... semi-stable... with Flash. Your 8.0.1 make NO ONE issue with Flash. Flash last release or last beta of v11.

Pity. But remain functions and speed of your 9.0: Divinely.

p.s.:
Without container Fx prozess blowing up to 500MB and more. Like container, when its active. But without container this story is more stable.

Otherwise no one issue with 9.0tete. Except Flash, perfect build.

Hmm... There is as yet no sign of crash of Flash in my system.
But if I can get the crush and detect the cause of the problem, I would like to fix it.

It's strange. Why my tete009 build didn't show the information in the Error Console...

@tete
And exactly therefore I find free wild betas of your builds, and not only finals, in order

Thank you for all the information.

I have uploaded the new build. I tried disabling PGO of browsercomps.dll.

I'm not sure whether this can fix the problem because the problem has not occurred in my system.

What's does exactly browsercomps.dll do ?

Still the same, tete :(

Hmm... It stumps me.

Error Console: (softpedia.com/rss)

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Warning: Error in parsing value for 'min-width'. Declaration dropped.
Source File: chrome://browser/content/browser.xul
Line: 0
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

I removed browsercomps.dll from my build, but RSS feeds work. Hmm.

I'm using the second build on my win7 x64 machine and RSS works fine. Haven't tried the third build though.

I got the below message when opening RSS, but it still loads fine.

Warning: Use of getAttributeNodeNS() is deprecated. Use getAttributeNS() instead.
Source File: http://www.softpedia.com/rss
Line: 0

I have confirmed the problem on 32-bit Windows XP SP3 on Virtual Machine.I will examine the cause of the problem.
http://samlab.ws/rss/

Problem confirm @win7 x86
Seems x86 systems are affected.
no error in error console.

Probably I found the cause of this problem. I will update my build later.

Sorry, please forget what I just said.

Probably I found the cause of this problem. I will update my build later.

When I copied the following files into the components folder, browsercomps.dll was loaded successfully and the trouble fixed.

Microsoft.VC80.CRT.manifest
msvcm80.dll
msvcp80.dll
msvcr80.dll

I suspect that this problem is caused by the source code of Mozilla.

Yea, it works.
If I encounter the similar problems when building with VC2010, I will also use the method.

Works for me , too !!!

manually changed version string in the manifest to version="8.0.50727.762".
now Tete's browsercomps.dll works

Yes in my Win7 SP1 only this exists
8.0.50727.762
8.0.50727.4940

this http://www.microsoft.com/download/en/details.aspx?id=26347 can help

Visual C++ 2005 Service Pack 1 Redistributable Package MFC Security Update

I do not know whether it is a final build for Firefox 9.0 release or not, but I am just building it now.

I had created prototype builds of Aurora 9.0b6 for several days, but a few illegal codes had been generated in the mozjs.dll, so I excluded some source files from PGO.

Unless another big problem occurs, I would like to upload my Aurora 9.0 candidates build1 in several hours.

tete dont make you crazy with new-versions-lechery of users...

I be happy only about your builds, once this have full features set of a tete build.
Only that, thus a patches and your PGO, make a fabulous full tete speed.

Is this a source for your PGO problems?
http://blog.mozilla.com/futurereleases/2011/11/10/type-inference-to-firefox-beta/

Hi Tete

The new FF 9.0 build1 won't start on my system. Crashes immediately.

My PC: Athlon 64 x2 Win7 32

Me too, crashy ! XP SP2 : http://tinypaste.com/4b71fcef

> Maybe fixed the startup problem by circumstances.

yes, thanks

Sorry, I have just uploaded the new build. I tried to fix the manifests of my build. I hope this could fix the startup problem.

I was faster .
Maybe because of the blazing fast new version. Now I have 310ms in Sunspider. With prev. version result was 340ms.

OMG rss in this version is broken
i have to downgrade i think

Hi, tete

Your first FF9 build is broken and the second FF9 build is good, but the patches are the same.

Is the first build 's failure not the patches' issue?

@tico-tico
Could you let me know where the problem occurs?

@xunxun
Yes, my first build had the wrong manifest data about OpenMP.
Both are the same patches.

Could you let me know where the problem occurs?

all rss feeds are broken

for ex.

http://www.softpedia.com/rss
http://www.kvraudio.com/rss/kvr_news_top.rss

official version works for me

RSS in FF9 (from Mozilla) broken too !

RSS in FF9 (from Mozilla) broken too !

Are you sure? Just tested on my virtual machines (WinXp SP3 and Win7 SP1 64-bit). Official version works fine with rss.

@JS
Yes, new added problems caused by PGO are in 3 source files which are the source code of Method JIT.

Is this a source for your PGO problems?
http://blog.mozilla.com/futurereleases/2011/11/10/type-inference-to-firefox-beta/

@tico-tico
In my environment, the RSS feeds seems to work with my last build. Does the Error Console show any message that is relevant to this problem?

My RSS is also broken using tete last build. Error console didn't show any information. The symptom is that Firefox can't recognize RSS feeds link, then open a dialog to ask the application to open with.

@tico-tico : Sure, FF 8 works fine for me (Win XP SP2)

[continued] Or will open a tab to show its xml.

Does the Error Console show any message that is relevant to this problem?

Only during FF start.

Failed to load native module at path 'c:\aurora-9.0-2011121800.en-US.win32-tete009-sse2-pgo\components\browsercomps.dll': (80004003) error 14001

Could not read chrome manifest file 'c:\aurora-9.0-2011121800.en-US.win32-tete009-sse2-pgo\extensions\{972ce4c6-7e08-4474-a285-3208198ce6fd}\chrome.manifest'.

Oups.. It seems like official Fx have the same problem. Maybe my urlclassifier3.sqlite is broken. Sorry for bothering you.

And after more test with Tb8 and Addons and themes, the please-request on this place:

Please, please build the Tb 3.1.x up to his EOL. Please...

I would like to build Extended Support Release (ESR) of Firefox/Thunderbird at present.

I don't know when old mozilla releases such as FF3.6.x and tb3.1.x are ended. I want to build the final edition of FF3.6.x or tb3.1.x

Dezember 2011 or April 2012.

LOL %-) sorry -> DeCember ;)

Hi

Well... ESR versions ist not a solution, at the MOMENT. Looking on Mozillazine (only for example, the biggest Tb forum in my area talk the same language ;) ) updating/upgrading/migrating Tb seems at times to be MORE tricky and probelmatic than biggest problems with migrating Fx.

I can sing to you a more hours song of my migrating from Tb2 to ("only") Tb3.1...

Therefore my please request to make new Tb 3.1 versions while MozMail do it.

It is impossible for me to support the 3 or more versions of Firefox or Thunderbird in the same period, because of a heavy burden to me.

Really pity.

But, on another hand, I must say the 3.1 changes are very lame...(?)
https://bugzilla.mozilla.org/buglist.cgi?field0-0-0=cf_status_thunderbird31;type0-0-1=equals;field0-0-1=cf_status_192;query_format=advanced;value0-0-1=.25-fixed;type0-0-0=equals;value0-0-0=.17-fixed

@Pik
Building Firefox/Thunderbird is not very hard, and if you have other edition's need, you can build it yourself. (For me, VC PGO build costs too long time )

@TeTe
Jvalue is changed on Firefox 9, so your some patches on it are not compatible, you can modify it.
I will build it after you land in next release 1-3 days.

How, long? For a build one in two or three months...

Its not a coincidence than tete's builds running BEST of all 3rd part builds for Win.

Cool :) WebGL as "own" lib.
http://groups.google.com/group/mozilla.dev.platform/browse_thread/thread/c0b1d383f9a8ca63

I have often been harassed by the same problem of PGO too...

@TeTe
You can use Editbin to enable largeaddress, then can excess 3GB memory.

@xunxun
I know 32-bit VS2005 tools have IMAGE_FILE_LARGE_ADDRESS_AWARE in the process header. When I was using 32-bit Windows, I had used 4GT RAM Tuning to build my builds.

Now I have built my builds on 64-bit Windows, so 32-bit VS2005 tools can handle 4GB addresses in theory.

That's right. WinXP must enable /3GB option boot. Now I use Win7 64bit to build Firefox.

But in theory, this only can handle 3GB memory, not 4GB. (This is Windows OS's shortcoming, and Linux can handle all 4GB but M$ didn't provide Linux-to-Windows cross compilers)

In 64-bit Windows, user-mode virtual address space for each 32-bit process which has IMAGE_FILE_LARGE_ADDRESS_AWARE flag is 4GB.
http://msdn.microsoft.com/en-us/library/windows/desktop/aa366778%28v=vs.85%29.aspx#memory_limits

Well, this maybe is my misunderstanding for PAE (64 bit OS seems not to use it).
In other word, in 64 bit Win OS, 32bit process can handle all 4GB memory, but Win Kernel don't keep any memory for the process, am I right?

32bit Win-kernel confiscated 2Gb OR 1Gb wiht /3GB switch in boot.ini.

32bit prozess can "hold" 4Gb from Win. But on 32bit Win, you see, it can "hold" only 2Gb or 3Gb (with say to OS, kernel becomes only 1Gb *and* the prozess have/set LAA bit).

On 64bit Win, 32bit prozess CAN "hold" his completly 4Gb.

But kernel space is not only space for Win-kernel. This is windows kernel modules space (!) Also for parts of kernel module drivers (!) With much fat hardware in the box, 1Gb kernel ADDRESS space *can* be to small.

Hello,

It is normally expected that the Earlybird icons are displayed as follows.
http://tete009.up.seesaa.net/image/earlybird8.png

Equally shit like Linikai. Crappiest software idea on the planet. WHITE icons... After MinimizeToTray, I "cant" find it beside of clock. wtf... :-\

And IconChanger sadly dont work. In Windows properties of exe, with my changed icon. Ok. But in the taskbar (in tab) and after minimized to tray: both with old/original icon.

Can you not use ANY another icon than this crap?

p.s.:
Ok, I mean, I know this is Mozillas icon

I was hoping that it was just some build/compile option like we had for your thunderbird builds before. I think I was always the one who pointed the "not-so-good-looking" icon issue.

Oh well, it's cosmetic and I think I could bear with it.

Thanks.

o_O ?
This is not a question of good looking. The "new" Mozilla icons for 3rd parts are either invisible or only white specks on dark Win themes.

btw, 8.0.1 build 2 is out

http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/8.0.1-candidates/build2/win32/

The crash reporter has been disabled on my build, so gCrashReporter is not defined.
If some add-on refers gCrashReporter, I suggest that you ask author of the add-on to fix this matter.

The flash rendered correctly in my 8.0.1 profile compared with chrome 14, firefox 3.6.24, opera 11.52 and IE9

I guest there is something to do with the flash player or your profile.

I can see the title. All right.

Flash updated ! Correctly now ! Thanks !

Wtf is the changes set of 8.0.1? Blocking Roboform, lol?? Or you have new "sites pack" for profiling? Or VC Updates?

With my settings, 8.0.1 is *FELT* twice faster than 8.0
o_O O_o

With 6 Mbit/s most sites bounces on the screen before I can say "Boo"

Another anomaly of your builds, but these is SIMPLY GREAT!
Thanks Master!

> Respin
> Issues with the Roboform fix mean we'll need build2. Deleted firefox/releases/8.0.1 to get a start on cleaning the internal mirrors up (they only remove 10 files+dirs per sync).

It's a good idea to wait the official release, isn't it?

When I will official releases, I would not using tete's builds...

Thank you for the information.
I changed the strings from "Aurora 8.0.1" to "Aurora 8.0.1 build1".

FF 8.0.1 is meant only for LINUX, on Windows stays FF 8.0.

柴田さんこんにちは、ご使用くださりありがとうございます。

実はThunderbird私的ビルドも、tmemutil.iniでPortable=1を指定すれば、ポータブル化できるかもしれません。

tmemutil.iniの説明には「Portable (Firefoxのみ)」と書いてますが、テストやサポートが大変にならないようにそう書いてます。^^;

今ちょっと試したら、ポータブル化出来そうな感じがしました。ただFirefox私的ビルドと同様、動作しないアドオンがあるかもしれません。

Thunderbird私的ビルドのアップデート方法ですが、上書きしますと動作がおかしくなったことが過去にあったので、できればFirefox私的ビルドと同じ感じでやっていただければと思います。

返信ありがとうございます。
たしかに複数のビルドをサポートするのは大変ですね。
Thunderbird私的ビルドのポータブル化オプションについては、
自己責任の上利用させていただきます。

テテさん，柴田さん，こんにちは。

横から失礼します。
あくまで自己責任ですが，例えばPortableApps.comで配布されているThunderbird Portable Editionを利用すると，手っ取り早くだいたいのビルドにおいてポータブル化することができます。

その方法は，PortablApps.comのThunderbird Portableインストール先の・・・
1.ThunderbirdPortable\Appフォルダ内，Thunderbirdフォルダを削除し，テテさんビルドを解凍してフォルダ名をリネーム（Thunderbird）したうえでAppフォルダ内に保存する（つまり，Appフォルダ内にある公式ビルドのものであろうThunderbirdをテテさんビルドのものに置き換える）。
2.ThunderbirdPortableを起動する。

ご参考までに

アイヴァーンさん、こんにちは。詳しいご解説ありがとうございます。
PortableApps.comのビルドを利用して、他のビルドをポータブル化できるんですね。^^

テテさん，こんにちは。

私的には非常に邪道な方法だとは思うのですが，これはMozilla製のものに限った構造ではなく，たとえばGoogle Chromeも同様に“GoogleChromePortable\App”フォルダ内のChrome-binフォルダを最新のものに取り換えると，勝手Portable Editionができてしまいます。
他には，同じGoogle Chrome Portable Edition（一応Dev版）を利用して，Chromium 勝手Portable Editionなんてものも可能です。

まあ，USBメモリなどに入れてPortable化するのでなければ，Mozilla ProductsやChromiumは起動オプションで問題ない（※）のでしょうが，手軽なのでちょっとお試し用にはコレを使ってしまっています

※たとえばFirefoxの場合には，
〜\firefox.exe -profile "プロファイルフォルダアドレス" -no-remote
といったようなショートカットを一つ用意すれば，簡単にできてしまいます。
おそらくChromiumでも同様の起動オプションはあると思うのですが，日本語の文章が少ない気がして調べていません（ぇ

P.S.最近，いつからかはっきり把握していませんが，公式ビルドもアーカイバーでインストーラーを見ると簡単に中身が抽出できるようになっていたのデスネ。ZIP版がリリースされなくても，問題なく公式ビルドをセットアップせずに利用できるようになっていました。

アイヴァーンさん、こんにちは。
Google Chrome Portableの中身をChromiumのものに置き換えて、ポータブル化できるのですね。普段使っている設定ごと持ち運べるのは、便利ですね。

テテさん、アイヴァーンさんこんにちは。
上記の返信でアイヴァーンさんに教えて頂いた、PortableApps版のThunderbirdを利用した方法で、こちらのビルドをポータブル化していましたが、
この度、ベータ版DLLでポータブル化オプションに対応されたということで、新たにインストールして、特に問題もなく重宝させていただいてます。
ポータブルデータの保存場所を指定するオプションについては、私はデフォルトのままで問題ないので使用していませんが、
こちらの方も、必要な方には重宝されるオプションだと思います。

柴田さんこんばんは。

PortableオプションがThunderbird私的ビルドでも動くことが分かりましたので、解説から「Firefox私的ビルドのみ」という記述を削除しました。

先日アップロードしたベータ版tmemutil.dllですが、一部のアンチウイルスソフトがトロイの木馬と判定します。Avira AntiVirを使用している環境では、私的ビルドの実行に失敗します。

Aviraに検体を送信しましたが、今のところトロイの木馬との判定は変わっていません。

そういえばベータ版tmemutil.dllを導入して、初めて起動したときはたしかに起動失敗しました。
それ以降は普通に使用出来ていたので、なにが原因なのか分かりませんでしたが、Avira AntiVirが原因だったんですね。
自分もAvira AntiVirを使ってます。原因が分かってすっきりしました。

検体を提出していたAviraから以下のメールが来ました。最新版の定義では、問題が解消されました。

ファイル 'xxxxxxxx.vir' は、'FALSE POSITIVE' と判定されました。つまり、このファイルはマルウェアではなく、誤って警告の対象となったものです。次回のいずれかの更新において、ウイルス定義ファイル (VDF) から検出用のデータが削除されます。

VirusTotalでも検出されなくなりました。ほっ。
http://www.virustotal.com/file-scan/report.html?id=c4e676d6d2780b78871c61cca239b3298ea7be1db26081923e08c742a831bc29-1323184327

Thank you for reporting the crash. I have confirmed it.
If the reason for the crash is identified, I will fix it.

If the crash has been caused by the bug of PGO, fixing the problem may take time.

Aurora 8.0 is unstable,crash several times,Where could find the old version(7.0)?

Temp fix: replace tete's xul.dll with the original one.

FF8 also crash on the page after I used VC2010 PGO to build it.

I have just updated my builds.
I specified the compiler option of content/media/nsMediaCache.cpp to -GL-.
I hope this could fix the problem.

Thanks! Works fine now.

Thanks tete for the quick response.

WORKS. Thx Massa.

New version fixed crash at paintball test in ietest drive too.

Thank you for reporting about this matter.
When a trouble occurs, it would help me a great deal if you could provide in detail about the trouble.

Firefox 3.6.24 has not been released. I would like to build my 3.6.24 after it is released.

I got omni.jar\chrome\browser\skin\classic\aero\browser\browser.css from the official build. Now all fine.

Thank you for using my build.
It seems that the design of the default theme was slightly changed.

It seems that the design of the default theme was slightly changed.

Yeah, but only for the non-official builds. Good for us we know how to change it back.
bwt Is omni.jar just a zip-file without any fancy things or not?

After having read the following document, I had a impression that omni.jar format may be different from normal ZIP format.
https://developer.mozilla.org/en/About_omni.jar

I think it's true only for the official builds.
Interesting thing: your omni.jar started with "PK" bytes (just like the standard zip), but in official "PK" stays with offset in 4 bytes. Is this a optimization?

I think I should try to optimize omni.jar.
Good article is here
http://www.raymond.cc/blog/archives/2011/04/26/edit-files-inside-firefox-4-omni-jar-to-auto-save-password/

black... (?) Its dark blue. And I like it B-)

I am building my Aurora 8.0 now. If build succeeds, I would like to upload it today.

Thank you in advance!

Hi Master

What is now _your_ "patch-extent" in 8.0? Compare with 7.0, 7.0.1, your TEST_versions of 7.0.1 and so on.

Now only SVG-multithreading? I lost overview...

Hello Pik, my Aurora 8.0 build includes the features of 7.0.1 TEST builds:

* Used the multithread processing for some SVG filters.
* Added a SSSE3 routine and a multithreaded routine in canvas putImageData().
* Added a multithreaded routine in canvas getImageData().
* Sped up the DrawTargetD2D::CreateGradientTexture().
* Used the prefetchnta instruction in some functions of js.
* Used the movdqu instruction in ClearValueRange().

We may be able to feel the effect of multithreaded-SVG filters at the following examples:

http://srufaculty.sru.edu/david.dailey/svg/Turbulence1.svg
http://srufaculty.sru.edu/david.dailey/svg/water3.svg
http://srufaculty.sru.edu/david.dailey/svg/scenes/popcorn12.svg
http://srufaculty.sru.edu/david.dailey/svg/newstuff/filterTurbulence10.svg
http://srufaculty.sru.edu/david.dailey/svg/scenes/popcorn11.svg
http://srufaculty.sru.edu/david.dailey/svg/matrixFilter.svg
http://srufaculty.sru.edu/david.dailey/svg/scenes/popcorn4.svg
http://srufaculty.sru.edu/david.dailey/svg/feMorphologyDilate.svg

Ok. It was only a generally question. Thanks.

p.s.:
But with svg, max. cpu load with "your" examples on my E8400 are 54%. Little bit to low for multithreading (?) :)

I think the Windows Task Manager shows the average CPU usage over a time interval.

When there are parts which run by single thread during a given time interval, I think the Task Manager will not show the CPU usage as 100%.

You must know Tete does not release builds except for the stables if you have been using his build for long time.

I know. I am using tete's builds for a long time B-) And sometimes, ex ante releases, a new version beheld daylight

Original Beta1 was blazing fast here O_O I think, I will slowly impatient

Tbind computes the virtual address of each function that is imported by my build's images (DLLs or EXEs) and write it into the Import Address Table (IAT) of my build's images.

Could you see the Remarks of BindImageEx?
http://msdn.microsoft.com/en-us/library/windows/desktop/ms679279%28v=vs.85%29.aspx

No, sorry :( I cant see in the Remarks why I cant find tbind in your Fx7 builds.

The tbind is available here:
http://www1.plala.or.jp/tete009/en-US/software.html#TBIND

Does this mean I should run tbind for your Aurora 8 build after installed?

Yes.
But when the effect cannot be actually felt, it need not necessarily be executed.

Yes, this build switches the software routine by CPU features.

It is an interesting change. I will try to think about applying the patch to my next builds.

I have apply TCP Optimizer (on XP) http://www.speedguide.net/downloads.php , but set in
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\AFD\Parameters
DefaultSendWindow and DefaultReceiveWindow to ("only") 64512.

With
user_pref("network.http.max-connections", 64);
user_pref("network.http.max-connections-per-server", 32);
user_pref("network.http.max-persistent-connections-per-server", 10);
user_pref("network.http.max-persistent-connections-per-proxy", 32);
I have at the moment no "hangers" or laggs. AS YET not :) But on some servers little slowdowns/hesitations (!) Fx7 2011100920

The 48 connects "sonic barrier" is therefore true O_O

Nope. 60 - 30 is HERE the barrier at hte moment. More make (64, 36 for example) dont make errors, but at downloads many sites...

PaleMoon is WITHOUT code changes (!) in Dromaeo (DOM) 19% to 25% faster than Mozilla. This are big number... How so? :)

If what is written in the document is true, I think it is great. But I do not know any details because I have not installed PaleMoon yet.

PaleMoon exists also as REAL portable. If you want simply compare Fx-version: its easy and timesaving.

p.s.:

but slower than other most of 3rd builds (tete, Ayakawa
and my self build)

Ayakawa? Ayakawa's VC10sp1 Fx 7.0.1 make 2925 in Dromaeo/DOM. tete's 2011100920 make 3579 Pure JS is 700 to 694. The golden age of Ayakawa was his 3.6

I don't know why Dromaeo/DOM test my editon is slower than tete009. In my here, my edition is 1757.33runs/s and tete009 is 1835.34runs/s.
But I also use tete009's patches (except tmemutil.dll though I can hack it ).
So I think I should enable Dromaeo/DOM PGO next editon.

ps: my edition's js test is faster than tete009 , this is the result between some FF7 editions:

sunspider
pcxfirefox : 281.5ms +/- 0.7%
tete009 : 288.3ms +/- 1.4%
ayakawa : 278.1ms +/- 1.0%

V8
pcxfirefox : Score: 3729
tete009 : Score: 3551
ayakawa : Score: 3433

Kraken
pcxfirefox : 7119.5ms +/- 0.7%
tete009 : 7209.9ms +/- 1.1%
ayakawa : 7200.9ms +/- 1.8%

peacekeeper
pcxfirefox : 4719 Points
tete009 : 4646 Points
ayakawa : 4572 Points

My main concern of PGO is to optimize for general use, so when I am collecting the profile data, I try to think the following tips as much as possible:
http://msdn.microsoft.com/en-us/library/aa289170%28VS.71%29.aspx#profileguidedoptimization_topic7

I think what we dont need, are PGO for pure benchsites

I like 5 :)
"If you have two scenarios that run for different amounts of time, but would like them to be weighted equally, you can use the weight switch (/merge:weight in pgomgr) on .PGC files to adjust them."

But not for "pure benchsites"... ;)

p.s.:
tete what your magic -Gx and -Ox options? %-)

I have set -O2 and -GT (and -GL) for compiler options. I don't know whether or not GT is helping to speed up my builds.

I didn't use -GT, and I may try this option next release build.

Ok. Why not simply /Ox /GL /Gr (not GR) and /Qimprecise_fwaits ? :-p

http://msdn.microsoft.com/en-us/library/19z1t1wy%28v=VS.90%29.aspx

Generally speaking, /Ox is not better than /O2, but cost more memory. /Gr will break function calls those are not _fastcall. About /Qimprecise_fwaits, I don't know this can or can't influence the code exception.

Ok...
http://stackoverflow.com/questions/5063334/what-is-the-difference-between-the-ox-and-o2-compiler-options

http://msdn.microsoft.com/en-us/library/59a3b321.aspx

http://forums.pcsx2.net/Thread-o2-vs-oX-optimize-vc2008

No knowledge without trials ;)
http://social.msdn.microsoft.com/Forums/en-US/vcgeneral/thread/5c6e5a54-06c1-4ba1-b355-a257195902bb

Okay. O2 is /Og /Oi /Ot /Oy /Ob2 /Gs /GF /Gy. That ought to do.

p.s.:
But what the hell is /GA?? o_O

About -Gr. When setting -GL, compiler can optimize the use of registers across function boundaries.
http://msdn.microsoft.com/en-us/library/0zza0de8%28v=VS.100%29.aspx

PGO, which include -GL, will further the optimization of register allocation, so there is no need to set -Gr.

About -GA. Using /GA for a DLL can result in bad code generation.
http://msdn.microsoft.com/en-us/library/yetwazx6.aspx

I don't think /Gr can be used with /GL.
Because /GL can decide the internal calling convention itself， /Gr will enforce fastcall.
/GA will only be effective for exe.

Funnily enough, MS write, /Gr is for example not compatible with /clr:safe or /clr:pure, but not with /GL.

For first, you oughta think, /GL and /Gr make against fight of war (??) :-\

No, /GL can decide the best internal calling convention, such as fastcall, stdcall, cdecl and thiscall. /Gr may overload his judgement, and may generate weird code (I think).
In addition, no one try to combine /GL and /Gr to build his projects (unless he want to convert all calls to fastcall really), at least I do not know.

May? "May" occurs from maybe and not from means (by all means) (?)

Ok. I dont will "sell" /Gr. That was pure inquisitiveness.

regards

There is no need to specify -Gr when -GL is specified.

The -Gr option forces compiler to use the ecx or edx register for the parameter passing of functions. This can cause adverse effects on compiler optimization when compiler must push the contents of the ecx or edx register to the stack before function call. The stack processing spends CPU cycles.

When -GL is specified, compiler can use the CPU registers for function call much more flexibly.

Not only that but also there are a couple of optimizations that are possible by using -GL. For more info, Please refer to the following document:
http://msdn.microsoft.com/en-us/library/0zza0de8%28v=VS.100%29.aspx

I have a little idea... Take 1:1 plugiin-container form Mozilla.

Thank you for your information.

On my system, I can install the Evernote Webclip into my Aurora 7.0.1 with portable mode.

Does the Error Console show some error message relevant to this issue?

After click install, 3 errors in the error consle

1. statse.webtrendslive.com : server does not support RFC 5746, see CVE-2009-3555
2. Warning: WARN addons.xpi: Ignoring duplicate locale in localized properties
Source File: resource:///modules/XPIProvider.jsm
Line: 569
3. Warning: WARN addons.xpi: Ignoring localized properties with no listed locales
Source File: resource:///modules/XPIProvider.jsm
Line: 577

Then, the mentioned error message popup,
Then, in the error console

Warning: WARN addons.xpi: Failed to install: [Exception... "Component returned failure code: 0x80520012 (NS_ERROR_FILE_NOT_FOUND) [nsIZipReader.extract]" nsresult: "0x80520012 (NS_ERROR_FILE_NOT_FOUND)" location: "JS frame :: resource:///modules/XPIProvider.jsm :: extractFiles :: line 988" data: no]
Source File: resource:///modules/XPIProvider.jsm
Line: 988

Thank you for all the information.

If you put my build into long path, this error may occur.
https://bugzilla.mozilla.org/show_bug.cgi?id=636268

In the Windows API, the maximum length for a path is MAX_PATH, which is defined as 260 characters.

Thanks for the suggestion.
It works now.

Please see and try the following reply.
http://support.mozilla.com/en-US/questions/876427

LOL. "Testenvironment" please...

It crosses just my mind, optimizations (including PGO data and with MSVC make with it), *must not* are *to strong* dependent with your CPU, tete :(

Test Environment: Windows XP Home Edition SP3 - Compaq Persario - AMD Sempron 3200+ 1.79 GHz 950 MB Ram - Windstream 3MB DSL. Speed measured using Lifehacker Stopwatch: http://lifehacker.com/396518/the-browser-stopwatch-speed+tests-page-load . Ran previous build first followed by latest build using same Mozilla Profile.

On my "apparatus" the new testversion is simply slower then previous test. Nothing much, but noticeable. Reasons:

1.
Your PC and line is maybe "slow enough" to conceal a difference.

2.
Page loadtime testsites deliver utterly different results depending on daytime... Or nighttime ;)

3.
*Your* CPU likes the newest changes. My does it sadly not.

I use tete's builds since years. Suchlike fluctuations between versions (also releases) comming ocassionally. Odd. Tcha...

Hey cool! But dont forget it, on *48%* WinPCs running XP and Canvas make 1.1% of Inet.

;)

(Hey guys, reminder. This is only my english.

NOPE :( Ok, I am a test-nut. With many keys and some superfast sites in the EU for my brain based slowmo-eye-speed-detection :D

From latest test builds, 2011100920 is and persists the fastest ("habitual" Inet with real ww-world sites).

The best bet is a fast "wild" around klicking on a one FAST site and also repeat previous klicks -> You might see which version /follow/ the klicks most rapidly.

But, Canvas running superb :)

Hi

Competitors. PaleMoon is WITHOUT code changes (!) in Dromaeo (DOM) 19% to 25% faster than Mozilla. This are big number... How so? :)

http://www.palemoon.org/technical.shtml#speed

Most of the real "slowness" I see has to do with the DOM tests (Dromaeo doesn't just test pure JS engine speed). Gecko is just plain slow at a lot DOM operations. I think that quickstubs and moving away from some XPCOM stuff was supposed to help with some of that. I remember at one point that Tracemonkey would bail out if you started doing DOM operations too, although I don't have any idea if that's been changed or not.

tete "we" have now PGO in Js or not?
https://bugzilla.mozilla.org/show_bug.cgi?id=605033

I have already applied PGO to my mozjs.dll.
I don't know whether problems actually occurred by enabling PGO.

p.s.:
For Canvas benches ;)

(klick on Stats in the anim window)
http://alteredqualia.com/canvasmol/#Curcumin

Again tested. I am dont talking about slowness. The builds are either way all fast.
But I am "downgrade" now still to 2011100920. This version is Fx in best form and with my Fx-config, I am beaten with your Fx Chrome16dev (strike!) :)

For Canvas benches ;)

(klick on Stats in the anim window)
http://alteredqualia.com/canvasmol/#Curcumin

Thanks for the link. I got very interesting results. 20fps with ENABLED hw acceleration and ~60fps with DISABLED hw acceleration. Something completely broken inside Fx hw code. Where art thou my darling bugzilla...

Also with originale Mozilla executables?

For comparing the speed of getImageData and putImageData, I think the following benchmark test is accurate.
bug519400.bugzilla.mozilla.org/attachment.cgi?id=426752

Enabling Azure slows down getImageData and putImageData.

Thank you very much for the interesting example.

For Canvas benches ;)
(klick on Stats in the anim window)
http://alteredqualia.com/canvasmol/#Curcumin

Also with originale Mozilla executables?

Yes. Someone already filled a similar bug. https://bugzilla.mozilla.org/show_bug.cgi?id=680800

I got very interesting results. 20fps with ENABLED hw acceleration and ~60fps with DISABLED hw acceleration.

Yesterday tested on the same hardware but with XP SP3 installed (previous was win7 home). Got > 100fps with hw acceleration enabled.
So it seems like win7-related (dx10?) issue.

Imho, with XP and hw acceleration enabled, hw acceleration is disabled ;) ;) ;)

It is, simply, hw acceleration issue of Fx ;)

I tried to slightly optimize my last TEST build by using the following sample. When enabling Direct2D acceleration, last TEST build has become a little faster than it was before.
http://alteredqualia.com/canvasmol/#Curcumin

yes, fps increased for me from 20 to 25. and sunspider result now is slightly better. i really like this build.

bwt. maybe it's a good idea to build test versions based on Fx8 beta to be sure that your new patches will work well with the incoming Firefox?

Fx 8.0b? GREAT :D

And guys, I mean it with "anomalies" really honest...

Good night :D

Sunspider faster? How much?

I would like to build TEST version with the source code of stable release, because it is easy for me to compare with existing build. Also I hope to test my builds on my free time I saved.

For your information, there are limits to what I can do to speed up JavaScript.

? Not nightlys, Betas or (better) RCs. That are existing builds B-)
Be unafraid of your community. With Betas/RCs/TEST(s) you are allowed making mistakes

The prefetchnta loads one cache line of data from address X to a L1 cache of processor.
_mm_prefetch:
http://msdn.microsoft.com/library/84szxsww%28v=VS.100%29.aspx

_mm_storeu_si128 (movdqu):
http://msdn.microsoft.com/library/w1k1k29t%28v=VS.100%29.aspx

Can it be, *theoretically*, that optimisations for low cpu load, occasionally, make the routine (self= slower? :)

It doesn't matter. But I would hope your patches can be compatible with 64bit FF.

Probably the patch attached with the TEST version supports x64.

In fact, I refer to those intrinsic functions such as __readfsdword. I think these functions can't be used on FF 64bit build.

I would like to fix my patches for x64 until I upload my next builds.

It's good news. I want to build FF 64bit on next release...

"Because I didn't know very much whether my patches of software prefetch reduce CPU time."

? You wrote, it do it (?) :) On AMD k7, AMD64, Penryn and "i"-series from Intel, it do it (here). Not testet on Atom (lol).

This is fewer speedup patch, but ok less cpu load patch.

p.s.:
I self dont need more JS power. I need more Peacekeeper "DOM" and "TextOperations" Power ;) ;)

It is not easy for me to optimize Firefox.
If I find the part of source code which I can optimize, I would like to try optimizing it.

This is quite understandably of course. This was not a bleating (!)

The crash was caused by Flashgot 1.3.3. Reverted to 1.3.2 and now it no longer crashes.

Thank you for the information.
I am using FlashGot 1.3.3. I have yet to get a crash.
If I can find out the cause of the crush in my build, I hope to fix it.

SSE2 makes things slower on AMD K8(Athlon 64 series) machines. I suggest you use SSE build.

thank for suggestion, but SSE2 build works faster for me. i have no habit of blindly choose.

I think additional inspection is necessary about this build. Thank you.

I think its all ok with Your TEST version. The sunspider is indeed worse by 0,1 ms or less... so lets forget that. In other tests like Kraken, V8 and Peacekeeper i got very slight increase in performance. I tested sse2 on win7 64x and intel core i3 mobile

It works and is slight faster, but "slight" is to little. I want more ;-)

@TeTe: will this inspection patch contain in your next release (FF8)?

Palemoon about:buildconfig :

- -enable-chrome-format=jar - -disable-mochitests - -enable-strip - -disable-necko-small-buffers - -disable-accessibility - -disable-parental-controls - -disable-windows-mobile-components

Tete, Could you consider adding some Configure arguments for speed ?

- -enable-chrome-format=jar - -disable-mochitests - -enable-strip - -disable-necko-small-buffers - -disable-accessibility - -disable-parental-controls - -disable-windows-mobile-components

These options can't increase js score.
I think tete may do that by changing his compiler to VC2010 Express and collect more js PGO information, and the latter can improve js test very significantly.

2 xunxun

VC2010 Express can do PGO? I always thought it's only for Premium/Ultimate version. Tell me please, where i can find more info about PGO in Express version.

@tico-tico

VC Express and higher edition compiler is the same except MFC and ATL. About PGO, this is related with VC linker. For some reasons, VC Express IDE has not the PGO GUI option, but the cl and link command can do PGO (you can type link /? to see if the link option has PGO option, if not, we can use higher edition linker), this may lack some PGO dlls (such as pgort100.dll), so you can get them from other edition.

Where I can find a big-technical-nerd-faithful release notes ;) for VC2010? PGO of 2010 is much more better than in 2008? o_O

I am still undecided as to whether to apply the test patches to my Aurora 8.0. I have tested the patches with my debug build on Core i5-2500K machine, but I think I need more test on other CPUs.

I have heard VS2010 Professional also can use PGO.

I have never tried to use PGO with VS2010 Express. Even if we can use PGO in Express Edition by bringing pgo files from somewhere, doesn't it constitute a violation of EULA?

but I think using PGO on Express version with files that you got on the internet is against Microsoft's licence agreement.
What's the point of separately selling the Express version and the normal version if you are allowed to use PGO with both?

Looks like I accitendally posted pretty much the same thing with what Tete just posted. Forget my post, please.

@Lux I agree with you perfectly.

Yea, this is only a thought. I have had VC genuine edition.

I will apply the patch of ClearValueRange() to my next builds.

I tested the new jsvalue.h.patch at V8 benchmark suite, and got a slight reduction in CPU time.

The CPU time percentage of JSObject::init(), on Core i5-2500K:
before: 4.81%, after: 1.35%

The CPU time percentage of JSObject::init(), on Athlon X2 5050e:
before: 4.95%, after: 2.78%

The above CPU time is the part where JSObject::init() calls ClearValueRange().

I may not apply the patch of software prefetch to my next builds.

Ok. But, why? o_O

Please create "plugins" folder as follows, and put the plugin in there.

For all profiles:
(My build's root dir)\plugins

For each profile:
(My build's root dir)\PortableData\AppData\Mozilla\Firefox\Profiles\xxxxxxxx.default\plugins

Easy way to open the Profile Directory:

1. Open about:support.
2. Push the "Open Containing Folder" button.

It works perfectly.
Thank you.

ご報告いただいて気づきましたが、確かに「名前を付けて保存」ウィンドウのパスが、ドメインごと(?)に変化しますね。

私もこの仕様は、余計なお節介だと感じました。オプションで変更できるようにした方が良かったのではと思います。

現時点では、このパッチを外すことはできます。しかし、かなり大がかりな変更なので、今後このパッチ周辺のソースコードが変更されていった場合、対応するのは困難そうです。

ですので申し訳ありませんが、現状のままビルドしていこうと思います。

レスありがとうございます。
そう、一言で表すなら正に「お節介」なんですよね。
この変更は特にリリースノート等でアナウンスされているわけでもないので、全くと言っていいほど知られていないようです。
何か変だなと思いつつ使っている人は多いんじゃないかと思います。
今後どうなるか分かりませんが、戻してもらえるよう本家にフィードバックを送って様子をみようと思います。

私はいつの間にか便利な機能がついたなあ、と重宝しておりました。
保存フォルダを記録するソフトを使っていたので、そちら側の機能だとばかり思っていましたがw
ここを読んで初めてFirefoxの機能だと気付きました。

ブラウザの使い方によって、この仕様変更が便利な場合もあれば、不便な場合もあるでしょうね。

私は、いつもほぼ同じ場所にファイルをダウンロードしているせいか、ご報告をいただくまで変更に気が付きませんでした。

人によっては、使い勝手に影響する変更だと思うので、振る舞いをカスタマイズするための手段はあった方が良かった気がします。少なくともabout:configには。

Since Firefox 6.0, the default value of "network.http.max-connections" has been increased from 30 to 256.
http://kb.mozillazine.org/Network.http.max-connections

It might be a good idea to decrease this value from 256.

Try disabling the network.prefetch-next in about:config

Thank you for using my builds.

I have already deleted my Aurora 6.0.2 from my PC, so I can not meet your request.

Thanks for your replying. Just wondering, from now on, is it possible for you to just keep one previous build after a new build is releasing?

Thanks anyway!

Because my website's free space is low, that is not possible...

Why dont you use Mediafire.com ?

Because I'll be glad if you could come by my site.

Have you ever considered to put your build to Google Code or create a project on SourceForge? I suppose you agree to publicize your method to build a private build for Firefox, which I do think so but have a hard to sort it out on your site, ^_^

btw, I do think it is important for a major player of 3rd part firefox building like youself to keep more than one version of your works.

Way to go, pal!

Firefox rules, tete009 rocks!

いつもお世話になってます。
本当に素晴らしい仕事です。

最新は今までのようにここから落とすのでいいとして
倉庫として例えばMediafireにこれまでのを置いといてもいいような…
でもサイトどころかPC自体からも削除しているんですか？ｗ
うーん勿体ない

@George
I'm familiar with my traditional style, so I want to keep it. I'm sorry to disappoint you.

@takaさん
私個人としては、古いバージョンはセキュリティーの脆弱性が含まれていることがあるので、最新版だけを公開しておけば良いかなと思っていました。
PCには過去のバージョンの節目のビルドはいくつか残してますが、今回のは残してませんでした。^^;

This is not your fault, but new versions occasionally sucks. Not least by reasons with addons.

I don't think the only newest of newest versions are a good idea for any Firefoxtwekaer... Or must tweaking each new major.

In consideration of the compatibility, I am planning to build the builds of the Extended Support Release.

Then, probably I will stop building my SSE builds.

I think 2011 nobody will miss SSE-builds sorely...

Here's another link:
http://download02.mozilla.org/pub/firefox/releases/7.0.1/win32/en-US/

Only fix one bug : Bug 680802: Installing a new add-on at the same time as installing a new version of the app which uses new properties from install.rdf will fail. r=robstrong, a=LegNeato

I have just uploaded it.

Thank you!!
I downloaded it at once.

P.S.
Thunderbird 7.0.1 is released what means whether it is said that I am done as planned. Feel different from the Security Update whether it is not necessary to be upset particularly,; but ...

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
ありがとうございます！
早速ダウンロードしました。

追伸
予定通りというかなんというか，Thunderbird 7.0.1もリリースされていますね。セキュリティアップデートとは違うので，別段あわてる必要はないかとも思いますが・・・

どういたしまして。

ただ今、Thunderbird 7.0.1をビルドしてます。ビルド中に問題が起きなければ、昼頃にはできると思います。

pymakeというものを使って、パラレルビルド中。少しはビルド時間を減らせそうです。

どういたしまして。Aurora 7.0 SSE2版の不具合を直すのに手間取って、公開が予定より少し遅れました。

卵がひび割れてくちばしだけ出てますね。Firefoxのauroraチャンネルとかのビルドにも、何かキャラクターを付けてくれないかなと期待しているのですが、今のところ何もないようで。^^;

BTW, hi, all, can you follow or votes the thunderbird bug :
https://bugzilla.mozilla.org/show_bug.cgi?id=653342 ??
The issue can exist in the Chinese character, but I think this also can exist in the Japanese character.
Thanks.

テテさんビルドのだけでもムササビ付けましょう！

ムササビは英語でflying squirrelなんですってね。♪?(?ε?)

I have uploaded my builds. Thank you for waiting.

いつもありがとうございます!

xD thank so much!!! DOWNLOADED hehe, now: 1h05 A.M (Vietnam).
Kamikize is my administrator ^^, our forum: http://tangtocnet.com/ have fun Lol

I have uploaded my builds. Thank you for waiting.

Show me the link ?

My unofficial builds are on the following URL:
http://www1.plala.or.jp/tete009/en-US/software.html#FFDL

I have built Firefox7 X86 using your patches.
http://pcxfirefox.googlecode.com/files/Firefox-7.0-en-pcx-win32-110928-vc2010-sse2-betterpgo.7z

Thanks.

The pleasure is mine.

Thank you for the information. I would like to upload my builds before the end of today.

Thanks, tete !

To apply my patch to the source code, you will need to rewrite the environment variables of the first few lines before running the apply.bat.

I have not added any speed tweaks to my configuration files.

I have some questions about the variables:
Which dir is the Source dir? Is it "%appdata%\Mozilla\Firefox"?

And whih dir is the patch dir? The one where the firefox.exe is located? Or a Sub-dir?

And the dry-run.bat is it to remove all patches or has it another effect?

Thanking you in anticipation

The patch files contained in my build's Source folder have been already applied to my build.

So when using my build, there is no need to think about those files.

Oh, do not use this tweak, it just modify some options in about:config and it is outdated, maybe slow down Firefox and not suit for all hardware/network: http://forums.mozillazine.org/viewtopic.php?t=53650

Default setting is the best, at least set content.notify.backoffcount;5 and content.notify.interval;500000.
No need change content.max.tokenizing.time, it auto set to content.notify.interval x 3
More: http://www.techradar.com/news/software/applications/8-hacks-to-make-firefox-ridiculously-fast-468317
Really make Firefox faster but I think Tete's Firefox no need this tweak anyway. : "> ~~~~~

I "must" now change from ayakawa's 3.6.2x to tete's. Ayakawa have drop 3.6 support.

He have imho all tete patches. All ?
And more (?) Is it Ayakawa's VC and/or his compiler options? Ayakawa's pgo? Ayakawa's patch collection? I dont know. What I know, Ayakawa's 3.6 feels alas noticeable faster...

I want to quit building Namoroka 3.6.x and Lanikai 3.1.x in the near future.

This was not the question

I do not apply Ayakawa san's patches to my builds now.

I have collected the profiling data for PGO from dozens of my frequently accessed sites manually.

I don't know nothing else about the other builder's builds.

I mean Ayakawa dont *write* patches. He "collect" only, make pgo and /contrive/ VC options. He is not a Captain, like you, but a good 1st officer ;)

I mean noticeable is a "exposition" of images by rendering of the site. Text (Netforums), JS and so on are not faster.

Text (Netforums), JS and so on are not faster.

But, also not slower...

With my english... o_O

@djkilla
The tweaks are different good for different cpu-speed and i-network speed. You can increase teh speed-feeling, but universally setting dont exists...

The endeavour of Mozilla for newly engines, is like Chromium, "dynamic" response of the engine on several varying computers, ever with a optimal speed.
The old, unaltered tweaks are long-since STALE. Also for 3.6.

The only one "rule" are a maximize of network settings and pipelining/requests, but NOT to overstate this (!!) The settings are:
ser_pref("network.prefetch-next", true);
user_pref("network.http.pipelining", true);
user_pref("network.http.proxy.pipelining", true);
user_pref("network.http.pipelining.firstrequest", true);
user_pref("network.http.pipelining.maxrequests", 8);
user_pref("network.http.max-connections", 48);
user_pref("network.http.max-connections-per-server", 30);
user_pref("network.http.max-persistent-connections-per-server", 10);
user_pref("network.http.max-persistent-connections-per-proxy", 30);
user_pref("network.http.request.max-start-delay", 5);

The settings of content.notify.interval, content.switch.threshold, content.max.tokenizing.time, nglayout.initialpaint.delay and content.notify.backoffcount are in 3.6 VERY STRONG DEPEDNING from i-network speed AND cpu-speed. And, from each other.
The importance of this settings receding in importance as Fx7 (defaults are not great noticeable badder)

For example (depednig), interval and threashold must be equal and tokenizing multiple integer of this. Initialpaint must be "good finite" number of interval/threashold. Best coefficient between tokenizing and interval/threshold are for me 4. 220 interval / 176 initialpaint = 1.25.

For example, my settings for tete's 3.6:
user_pref("content.max.tokenizing.time", 880000);
user_pref("content.notify.interval", 220000);
user_pref("content.switch.threshold", 220000);
user_pref("content.notify.backoffcount", 2);
user_pref("nglayout.initialpaint.delay", 176);

All my settings on 3.85Ghz E8400, 6 Mibt/s AND Fx 3.6:
user_pref("browser.history_expire_sites", 2000);
user_pref("browser.cache.memory.capacity", 65536);
user_pref("browser.cache.disk.capacity", 76800);
user_pref("browser.sessionstore.interval", 30000);
user_pref("content.notify.ontimer", true);
user_pref("content.interrupt.parsing", true);
user_pref("content.max.tokenizing.time", 880000);
user_pref("content.notify.interval", 220000);
user_pref("content.switch.threshold", 220000);
user_pref("content.notify.backoffcount", 2);
user_pref("dom.disable_window_status_change", true);
user_pref("network.dnsCacheEntries", 512);
user_pref("network.dnsCacheExpiration", 3600);
user_pref("network.prefetch-next", true);
user_pref("network.http.pipelining", true);
user_pref("network.http.proxy.pipelining", true);
user_pref("network.http.pipelining.firstrequest", true);
user_pref("network.http.pipelining.maxrequests", 8);
user_pref("network.http.max-connections", 48);
user_pref("network.http.max-connections-per-server", 30);
user_pref("network.http.max-persistent-connections-per-server", 10);
user_pref("network.http.max-persistent-connections-per-proxy", 30);
user_pref("network.http.keep-alive.timeout", true);
user_pref("network.http.keep-alive.timeout", 30);
user_pref("network.http.request.max-start-delay", 5);
user_pref("network.http.connect.timeout", 30);
user_pref("ui.submenuDelay", 40);
user_pref("nglayout.initialpaint.delay", 176);

I haven't used pale moon edition, so I have no idea...

I tried Palemoon. If you put it to test, Palemoon is somewhat more stable but slower in both rendering Direct2D/Direct3D canvas and multithread

Sorry, I see the question has already been answered... hope you reconsider though, maybe after a few more builds perhaps?

You can wait for FF7 release, because tete only build the release edition.8-)

I have seen him release a few early builds, but still... it's understandable, especially now with the "accelerated release schedule" it'll be harder to stay on top of it all.

Does PcX build optimized Firefox 7.0b2?;-)

@Realgeorge
No, I have more important work to do. So I only build the latest release in my spare time as TeTe's. Thanks for the support.

I don't know about that...

On Pentium M PC, I have heard that SSE version is faster than SSE2 version.

so for YOUR PC, yr using SSE2 version right?

I am using SSE2 version on my Sandy Bridge (Core-i5 2500K) machine.

On Intel Atom or AMD K8, I have also heard that SSE version is faster than SSE2 version.

on amd64 x2 dual core 4400+ what version will faster?

@justin
On AMD K8, perhaps SSE version is faster than SSE2 version.

@TeTe
BTW, why don't you build thunderbird with SSE2?
I build with SSE2: http://pcxfirefox.googlecode.com/files/thunderbird-6.0-pcx-win32-110817-vc2010-sse2-pgo.7z
After testing it, it runs well.

benchmark tes sunspider i compare 3 versions firefox 6

============================================
RESULTS (means and 95% confidence intervals)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - PCX firefox
Total: 152.9ms +/- 1.2%
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
============================================
RESULTS (means and 95% confidence intervals)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -teteh firefox
Total: 156.6ms +/- 1.1%
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
============================================
RESULTS (means and 95% confidence intervals)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - normal firefox
Total: 178.2ms +/- 1.0%
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

pcx firefox in this test sunpider java benchmark is more faster all Both version teteh and pcx using version SSE2
my computer is core 2 quad Q9550 with 4 ghz

@PcX
It is simply a measure for cutting down on labor.

Thank you for waiting. I have uploaded my builds of Aurora 6.0 (Firefox 6.0).

お待たせしました。先ほどAurora 6.0 (Firefox 6.0)私的ビルドをアップロードしました。

いつもありがとうございます

gfx.font_rendering.directwrite.enabled を true にし、
gfx.font_rendering.cleartype_params.〜
の各設定をいじると、フォントが滑らかに表示されるようになりましたね
ｷﾚｲ*･゜･*:.｡..｡.:*･゜(n‘∀‘)ηﾟ･*:.｡. .｡.:*･゜･*

thx for new build!

It's my pleasure!
どういたしまして!

ありがとうございます。
早速ダウンロードさせていただきました。

Thank you.
I was going to download it immediately.

こんばんはです。情報ありがとうございます。m(_ _)m

うわ、16日頃リリースされるのかと思ったら、もうreleasesディレクトリにWindows版のソースコードがありますね。

これからビルドの準備に取りかかろうと思います。ただ、今日中にアップできるかは分からないです。

Waiting for your patches..

I have built firefox6.
link: http://pcxfirefox.googlecode.com/files/firefox-6.0-pcxfirefox-win32-110815-PGO.7z

Thanks for tete009's patches.

and what version is better now?

@PcX
You're welcome!

@justin
Mozilla recommends updating to the latest version, but some add-ons may not work with it.

Hello,
I do not have any plans to build the x64 version of Firefox for a while.

firefox 64 bits more slowly with x86 version only advanced is if you open 500 tab's and .exe firefox.exe run more 2gb in 64 bits process, x86 limited to 2 gb maximum per .exe

テテさん，こんにちは。
先 - http://cgi38.plala.or.jp/tete009/board.cgi?id=T1310507811.94#T1310507811.94 - にWindows版公式ビルドの自動更新では通知されないため，変更点がないと判断してビルドの計画がないとのこと。ただ，いつのまにかwww.mozilla.comおよびmozilla.jpで配布されているものは5.0.1 or 3.6.19になっていますね。

しかしながら，これはMozilla JapanのBlogにあるように便宜上ビルドされて公開されているにすぎないようで，更新の必要性はないと明言されています。

[Mozilla Japan Blog：Firefox と Thunderbird の Mac OS X 10.7 Lion 対応状況から引用] - 13/07/2011 - http://mozilla.jp/blog/entry/7013/
昨日、Lion との互換性を向上させた Firefox 5 と 3.6 のアップデート (バージョン 5.0.1、3.6.19) を公開しました。現在 Mac で Firefox をお使いの皆さんには、製品の自動更新機能を通じて 48 時間以内にこれらのアップデートをお届けします。リリースの都合上 Windows 版と Linux 版も用意していますが、Mac 版以外での変更点はなく、更新の必要はありません。

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
The renewal of Firefox 5 and 3.6 that improved interchangeability with Lion (versions 5.0.1 and 3.6.19) was opened to the public yesterday. We will send these updates to you who use Firefox with Mac through the automatic update function of the product within 48 hours now. It is not, and is not necessary to update the change point excluding the Mac version though the Windows version and the Linux version are prepared for the convenience of release.

自動更新でも更新されず，特に更新点もないにもかかわらず公式ビルドはアップデートされ，それでもなおユーザーには特に更新を求めないというのもちょっと中途半端な対応に感じられてしまいます。
まあ，ユーザーの好きにしてちょ！ってことかもしれませんけど