Broken FF 3.6b4 build

[Toan Pham]

Hi,

today, i checked out FF source from:

http://ftp.mozilla.org/pub/mozilla.org/firefox/releases/3.6b4.source.tar.bz2

and attempted to build. The build failed while compiling nsDocument.cpp
because it couldn't find include file 'nsMediaDecoder.h'.
I then added include path '/content/media' to make file /content/base/Makefile
and everything worked.

I would like to know who maintains/verifies FF lastest beta releases.
It would be nice if we can fix it in the repo.

thanks,

Toan

[Ted Mielczarek]

Nobody maintains the source tarballs, they're just produced as a part of our
release process. I'm quite sure that the code from the repo is ok (because
the release was built from there), but it's possible there's a problem with
the source tarball, or your build environment that caused your build to
fail.

-Ted

[Toan Pham]

Hi all, after updating firefox, i ran it on my system, and i got an error:

run-mozilla.sh: 131 : 5265 illegal instruction

and so, firefox never launch. I googled it and found that this
typically happen when the firefox subshell died unexpectedly.
Has anyone experienced this before? The system i am running on is:

OS: Linux 2.6.21
CPU: MSTI PDX-600 - 1GHZ (i486 compatible)
RAM: 256 MB

Thank you,

Toan

[Benjamin Smedberg]

On 12/4/09 10:33 AM, Toan Pham wrote:

> OS: Linux 2.6.21
> CPU: MSTI PDX-600 - 1GHZ (i486 compatible)
> RAM: 256 MB

I don't know much about the PDX-600, but if you break into a debugger on the
crash (launching with firefox -g for example), what is the illegal
instruction? Is it possible that the processor claims support for SSE2 or
the CMOV instruction but doesn't actually implement them? It's also possible
that we're misdetecting support for SSE2 or similar instruction extensions,
either in core code or in nanojit.

--BDS

[Benjamin Smedberg]

On 12/4/09 10:52 AM, Benjamin Smedberg wrote:

> I don't know much about the PDX-600, but if you break into a debugger on
> the crash (launching with firefox -g for example), what is the illegal
> instruction? Is it possible that the processor claims support for SSE2
> or the CMOV instruction but doesn't actually implement them? It's also
> possible that we're misdetecting support for SSE2 or similar instruction
> extensions, either in core code or in nanojit.

Also, if you could paste the contents of /proc/cpuinfo that may help narrow
the problem.

--BDS

[Toan Pham]

Benjamin,

You're absolutely right. There must had been invalid instructions
executed on the PDX-600 platform.
I tested the build OS image in qemu and firefox launched just fine.
Because my OS image is a linux OS that was cross-built from Scratch,
it does not have gdb debugger
in there yet. In the meantime, i'll work on debugging utils and
report back to you as to what instruction
it crashed.

Thanks

CPU INFO:

processor : 0
vendor_id : Vortex86 SoC
cpu family : 5
model : 2
model name : 05/02
stepping : 2
cpu MHz : 1000.072
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu tsc cx8
bogomips : 2007.74
clflush size : 32

> _______________________________________________
> dev-builds mailing list
> dev-b...@lists.mozilla.org
> https://lists.mozilla.org/listinfo/dev-builds
>

[xxx xxx]

On Dec 3, 10:25 pm, Toan Pham <tpham3...@gmail.com> wrote:
> Hi,
>
> today, i checked out FF source from:
>

> http://ftp.mozilla.org/pub/mozilla.org/firefox/releases/3.6b4.source....

>
> and attempted to build.  The build failed while compiling nsDocument.cpp
> because it couldn't find include file 'nsMediaDecoder.h'.
> I then added include path '/content/media' to make file /content/base/Makefile
> and everything worked.
>
> I would like to know who maintains/verifies FF lastest beta releases.
> It would be nice if we can fix it in the repo.
>
> thanks,
>
> Toan

I tried to build firefox but did not have alsa lib installed so i used
--disable-ogg and --disable-wave
building failed with the same error so i installed alsa lib an removed
this two options building was ok