Can you put the about:buildconfig output for the mozilla.org package over here? Maybe we missed something in our build.
Also note, our build uses the pango and cairo exclusively for text rendering, while the mozilla build uses xft to render text. Pango+cairo is still slightly slower on X.org 6.8, but will speed up a lot with 6.9. Also, the gnomies are busy optimizing pango for speed at this moment (IMHO firefox 1.5 isn't very slow at that page, but I'm running a pango devel snapshot at this machine)

The config is below. I'm running Xorg 6.9 myself, so I expect it's the cairo rendering that makes the difference. Is cairo rendering needed? Does it have any advantage for firefox text rendering at the moment?

about:buildconfig

Build platform
target
i686-pc-linux-gnu

Build tools
Compiler Version Compiler flags
gcc gcc version 3.3.2 20031022 (Red Hat Linux 3.3.2-1) -Wall -W -Wno-unused -Wpointer-arith -Wcast-align -Wno-long-long -pedantic -pthread -pipe
c++ gcc version 3.3.2 20031022 (Red Hat Linux 3.3.2-1) -fno-rtti -fno-exceptions -Wall -Wconversion -Wpointer-arith -Wcast-align -Woverloaded-virtual -Wsynth -Wno-ctor-dtor-privacy -Wno-non-virtual-dtor -Wno-long-long -pedantic -fshort-wchar -pthread -pipe -I/usr/X11R6/include

Configure arguments
--enable-application=browser --enable-update-channel=release --enable-update-packaging --disable-debug '--enable-optimize=-Os -freorder-blocks -fno-reorder-functions -gstabs+' --disable-tests --enable-official-branding --enable-default-toolkit=gtk2 --enable-xft --disable-freetype2 --enable-svg --enable-canvas --enable-static --disable-shared

Hi,

Well, I've had similar issues with Arch's testing mozilla-firefox 1.5 package myself. It was redering things slowly, the smooth-scrolling was unusable sometimes (especially when scrolling large <table> elements). Also the browser sometimes crashed. Usually it was on Wikipedia pages.

Before, I've tried Shadowhand's firefox-devel 1.5 (stable) build and it didn't have such issues. But the firefox-devel PKGBUILD didn't have the correct icon (I can't live with the ugly X ;P)

So I've decided to compile my own mozilla-firefox 1.5 testing package. At first I didn't touch the PKGBUILD or mozconfig at all. The results were the same as with the precompiled package. Same slowness, crashed once.

So I've decided to remove all the extra patches the testing PKGBUILD applies to the sources (I left launcher.patch only). Also I modified the mozconfig a bit (copied the Shadowhand's one, but they don't differ dramatically). And guess what? The problems are over! Everything is snappy, smooth-scrolling works great. The page you've posted above gives me only 40% CPU usage (Athlon 2500+) when scrolling, while with the Arch's packeg it's just painfully slow.

I guess the patches are faulty here, but I'm no expert.

My mozconfig:
# load defaults from src tarball
. $topsrcdir/browser/config/mozconfig
# add our own options
ac_add_options --prefix=/opt/mozilla
ac_add_options --with-default-mozilla-five-home=/opt/mozilla/lib/firefox
ac_add_options --enable-official-branding
ac_add_options --enable-crypto
ac_add_options --enable-optimize="#CFLAGS#"
ac_add_options --disable-debug
ac_add_options --disable-tests
ac_add_options --disable-logging
ac_add_options --disable-installer
ac_add_options --disable-activex
ac_add_options --disable-activex-scripting
ac_add_options --enable-strip
ac_add_options --enable-xft
ac_add_options --enable-toolkit-gtk2
ac_add_options --enable-default-toolkit=gtk2
ac_add_options --enable-system-cairo
ac_add_options --enable-extensions=cookie,xml-rpc,xmlextras,pref,transformiix,universalchardet,webservices,inspector
ac_add_options --disable-toolkit-xlib
ac_add_options --disable-toolkit-qt
ac_add_options --disable-toolkit-gtk
ac_add_options --disable-freetype2
ac_add_options --disable-pedantic
ac_add_options --with-system-jpeg
ac_add_options --with-system-zlib
ac_add_options --with-system-png
ac_add_options --with-system-mng
ac_add_options --enable-svg
ac_add_options --with-pthreads
ac_add_options --disable-mailnews
ac_add_options --disable-calendar
ac_add_options --disable-composer
ac_add_options --enable-single-profile
ac_add_options --disable-profilesharing
ac_add_options --disable-gnomevfs
ac_add_options --disable-xinerama
#ac_add_options --disable-mathml

I too have had a problem with Firefox from Testing's performance, it is noticeably slower on rendering and handling of certain things (like javascript's ScrollBy function) than Shadowhand's package. On the other hand, it has indeed fixed a few bugs that I had been running into (problems with right-click and Save As, for instance), so it's a bettersweet package ;)

about:buildconfig

Build platform
target
i686-pc-linux-gnu

Build tools
Compiler Version Compiler flags
gcc gcc version 4.0.3 20051006 (prerelease) -Wall -W -Wno-unused -Wpointer-arith -Wcast-align -Wno-long-long -pedantic -march=i686 -O2 -pipe -pthread -pipe
c++ gcc version 4.0.3 20051006 (prerelease) -fno-rtti -fno-exceptions -Wall -Wconversion -Wpointer-arith -Wcast-align -Woverloaded-virtual -Wsynth -Wno-ctor-dtor-privacy -Wno-non-virtual-dtor -Wno-long-long -pedantic -march=i686 -O2 -pipe -fshort-wchar -pthread -pipe -I/usr/X11R6/include

Configure arguments
--enable-application=browser --with-system-nspr --with-system-jpeg --with-system-zlib --with-system-png --with-system-mng --with-pthreads --disable-tests --disable-debug --disable-installer '--enable-optimize=-march=i686 -Os -pipe' --disable-xinerama --enable-default-toolkit=gtk2 --enable-official-branding --disable-xprint --enable-strip --enable-pango --enable-system-cairo --enable-svg --enable-canvas --prefix=/opt/mozilla --with-default-mozilla-five-home=/opt/mozilla/lib/firefox --enable-crypto --enable-extensions=cookie,xml-rpc,xmlextras,pref,transformiix,universalchardet,webservices,inspector,typeaheadfind --enable-single-profile --disable-profilesharing --disable-gnomevfs

I guess omitting --enable-xft hardcodes firefox to pango. When pango doesn't render things, fonts look very bad with firefox in testing. I will add xft support to the next package update.

why is -Os forced in this kind of ugly anyway? what if i have -01 in my CFLAGS? :)
not that this is a laughing matter of course, just wondering why firefox is built optimized for size.

Oh, I like pango+cairo way, it solve CJK bold font problem (if using libxft, we CJK user have to patch libxft to get CJK font bold display). Seems a double-edge sword currently, :-)

My previous build (the one I described in my previous post) had some issues. Namely: it froze on some pages (the rendering of them took soem time, sometimes ~1min).

So I decided to compile firefox again, usign the PKGBUILD (and patches) from the TESTING tree, except for the disabling of firefox-nopangoxft.patch. Since I wondered if pango isn't the problem there, I decided to remove pango entirely.

My about:buildconfig
Compiler Version Compiler flags
gcc gcc version 4.1.0 20051112 (experimental) -Wall -W -Wno-unused -Wpointer-arith -Wcast-align -Wno-long-long -pedantic -march=athlon-xp -O2 -fomit-frame-pointer -pthread -pipe
c++ gcc version 4.1.0 20051112 (experimental) -fno-rtti -fno-exceptions -Wall -Wconversion -Wpointer-arith -Wcast-align -Woverloaded-virtual -Wsynth -Wno-ctor-dtor-privacy -Wno-non-virtual-dtor -Wno-long-long -pedantic -march=athlon-xp -O2 -fomit-frame-pointer -fshort-wchar -pthread -pipe -I/usr/X11R6/include

Configure arguments
--enable-application=browser --with-system-nspr --with-system-jpeg --with-system-zlib --with-system-png --with-system-mng --with-pthreads --disable-tests --disable-debug --disable-installer '--enable-optimize=-march=athlon-xp -Os -fomit-frame-pointer' --disable-xinerama --enable-default-toolkit=gtk2 --enable-official-branding --enable-xft --disable-pango --disable-xprint --enable-strip --enable-system-cairo --enable-svg --enable-canvas --enable-mathml --prefix=/opt/mozilla --with-default-mozilla-five-home=/opt/mozilla/lib/firefox --enable-crypto --enable-extensions=cookie,xml-rpc,xmlextras,pref,transformiix,universalchardet,webservices,inspector --enable-single-profile --disable-profilesharing --disable-gnomevfs

Also, could we have mathml enabled by default in the package?

I have only observed one major problem: http://www.imageshack.us (and only the front page!) causes Firefox to become slow and unresponsive, until the tab showing that site is closed.