aebrahim: Linux 2003-11-03 GCC 3.3.2 XFT/GTK2 Pentium4/SSE2

[aebrahim]

I've compiled Firebird on Linux using GCC 3.3.2 with XFT and GTK2 for the Pentium 4 with SSE2 optimisations. I compiled this build using Fedora Core 0.95. The build is here:

http://pryan.org/firebird/aebrahim/Mozi ... e2.tar.bz2 (8.4MB)

I used the following .mozconfig to build:

. $topsrcdir/browser/config/mozconfig
ac_add_options --disable-tests
ac_add_options --disable-debug
ac_add_options --enable-optimize="-O3 -march=pentium4 -msse2 -mfpmath=sse"
ac_add_options --enable-xft
ac_add_options --enable-default-toolkit=gtk2

My builds have the following preference set:
pref("general.useragent.vendorComment", "aebrahim");

[GNU/Ben]

$topsrcdir/browser/config/mozconfig
ac_add_options --disable-tests
ac_add_options --disable-debug
ac_add_options --enable-optimize="-O3 -march=pentium4 -msse2 -mfpmath=sse"
ac_add_options --enable-xft
ac_add_options --enable-default-toolkit=gtk2

I am going to make my own FB build. But I'm still not sure about all the options for .mozconfig. Since the build of yours that I used worked nice, I wanted to know, is that your entire .mozconfig?

[Yek]

I noticed a bug in my 3.3.2 builds (but on linux, and with specific gcc chip optimizations), and would like to see if its in your build. To get this to happen (or not to happen), open the javascript console and if there's messages (or not), try clicking the Clear button. If that works ok, then try entering an expression(1+1 will do) and click on Evaluate. By the way: Don't do this after you've loaded up your browser with 10 tabs of stuff you actually wanted to see :) If it occurs, it crashes the browser to desktop. Also: No extensions or themes were used.

Thanks much. I've gotten someone else to verify it in the bugs forum, but any other input might be helpful for the bug report to be able to accurately report optimizations that crash with 3.3.2. So far, it happens on gcc 3.3.2 builds optimized for either p4's or athlonxp's. Both of us who had the bug on 3.3.2 tried it on 3.2.3 and it didn't display that bug. It also was gone when we used 'official' optimizations but used 3.3.2.

[aebrahim]

$topsrcdir/browser/config/mozconfig
ac_add_options --disable-tests
ac_add_options --disable-debug
ac_add_options --enable-optimize="-O3 -march=pentium4 -msse2 -mfpmath=sse"
ac_add_options --enable-xft
ac_add_options --enable-default-toolkit=gtk2

I am going to make my own FB build. But I'm still not sure about all the options for .mozconfig. Since the build of yours that I used worked nice, I wanted to know, is that your entire .mozconfig?

Yep, that's my entire .mozconfig. Make sure that you have the leading dot-space before $topsrcdir as I have in my first post. Also note that this .mozconfig will only work once you have the tree pulled. To initially pull the tree, use this .mozconfig:

export MOZ_PHOENIX=1
mk_add_options MOZ_PHOENIX=1

and then do a make -f client.mk checkout

Then once the checkout is successful, replace that .mozconfig with the once I'm using.

[aebrahim]

I noticed a bug in my 3.3.2 builds (but on linux, and with specific gcc chip optimizations), and would like to see if its in your build. To get this to happen (or not to happen), open the javascript console and if there's messages (or not), try clicking the Clear button. If that works ok, then try entering an expression(1+1 will do) and click on Evaluate. By the way: Don't do this after you've loaded up your browser with 10 tabs of stuff you actually wanted to see If it occurs, it crashes the browser to desktop. Also: No extensions or themes were used.

Thanks much. I've gotten someone else to verify it in the bugs forum, but any other input might be helpful for the bug report to be able to accurately report optimizations that crash with 3.3.2. So far, it happens on gcc 3.3.2 builds optimized for either p4's or athlonxp's. Both of us who had the bug on 3.3.2 tried it on 3.2.3 and it didn't display that bug. It also was gone when we used 'official' optimizations but used 3.3.2.

Yup, I can confirm (using my 2003-11-04 build) a crash in the JS console when pressing on the clear button and also when trying to evaluate 1+1. I'm going to recompile using "-O2 -g" as my optimisation arguments and try this again and let you know in this thread how that turns out.

[Yek]

Thanks for testing that.

http://bugzilla.mozilla.org/show_bug.cgi?id=224731 is the bug I've filed for this. Feel free to add any comments there as well.

[aebrahim]

Thanks for testing that.

http://bugzilla.mozilla.org/show_bug.cgi?id=224731 is the bug I've filed for this. Feel free to add any comments there as well.

I've commented in the bug. Probably makes more sense to continue discussion of this issue in Bugzilla.