The unconditional pointer update approach is by no means atomic unless you use memory barriers or atomic instructions. There is a reason C++11 added <atomic>.

/Users/morten/Project/Sudoku> time sudoku worlds_hardest.sudoku
|8 1 2|7 5 3|6 4 9|
|9 4 3|6 8 2|1 7 5|
|6 7 5|4 9 1|2 8 3|
|1 5 4|2 3 7|8 9 6|
|3 6 9|8 4 5|7 2 1|
|2 8 7|1 6 9|5 3 4|
|5 2 1|9 7 4|3 6 8|
|4 3 8|5 2 6|9 1 7|
|7 9 6|3 1 8|4 5 2|

real 0m0.083s
user 0m0.073s
sys 0m0.003s

You're almost right, but the main problem is when a team gets obsessed with other issues than actually writing code. You want as little time as possible to be spent on debugging and rewriting. In order to achieve that you need some tools. To avoid rewriting you need some way of specifying what the software is supposed to do before actually making it. UML can be used for that, but it's not a goal in itself. To avoid endless debugging session you need tests, unit tests can be used, but I've found it far more productive to write code that has a lot of debugging code in it. Since I'm mainly writing C/C++, this will be in the form of asserts and #if DEBUG ... #endif, but the main idea is to catch errors as early as possible during program execution. In my experience it's far more productive to get dedicated testers to test end-user functionality and file bugs than to make the programmers who wrote the code write code that checks if the code they wrote did what they thought it should do. The reason is that most bugs occur because the programmers who wrote the code failed to consider some particular corner case when they wrote the code, and they will likewise fail to write a test to test it...

In conclusion, the primary goal is to develop a product, not to write tests, not to make specifications and not to make clean revision histories. However, when used right, specifications, testing and version control enables you to develop the product faster and with higher quality.

Yeah, C++ and Java is probably equivalent in the time it takes to solve the actual problem -- but then you spend 4 times as long debugging your C++ code because you have some weird write-past-the-end of an array or use-after-free bug that somehow works most of the time and when it goes wrong it only affects some completely unrelated piece of code from where the bug is that got it's data structures ruined. These classes of bugs can bring down novice programmers who can spend weeks trying to figure out where it's going wrong, and they just don't happen with Java or C# or Python.

That's just bullshit - compilers often get it wrong because of things like alias analysis and register spilling/rematerialization. Not to mention the really crappy autovectorization they do. If you really care about performance you use vector intrinsics and check the assembly output of the compiler for any obvious cockups.
There are also many programmers writing code for embedded procesors and DSPs that don't have compilers as mature as those for x86 and doesn't have any cycles to waste.

Javascript has closures and functions as first order objects, Java doesn't. Java has classes and inheritance, Javascript doesn't (but you can sort of implement it yourself). They are just completely different languages and need to be treated as such.

The 64-bit linker can link 32-bit object files, albeit without link-time code generation.

Here: http://msdn.microsoft.com/en-us/library/5wy54dk2(v=VS.80).aspx ... However it does not support link-time code generation when you're linking object files in a non-native format.

Ah, sorry -- I see the difference now, you're using link time code generation which the x64 bit linker is not capable of when the machine type is x86.

hello.c:

#include
#include

int main(int argc, char *argv[])
{
    printf("Hello world!\n");
    return EXIT_SUCCESS;
}

hello.bat:

@ECHO OFF
SETLOCAL
CALL "%VS80COMNTOOLS%..\..\vc\vcvarsall.bat" x86
cl /MT /c hello.c
ENDLOCAL
SETLOCAL
CALL "%VS80COMNTOOLS%..\..\vc\vcvarsall.bat" amd64
link /MACHINE:x86 /LIBPATH:"%VCINSTALLDIR%\lib" libcmt.lib hello.obj
ENDLOCAL

output:

C:\Project>hello.bat
Setting environment for using Microsoft Visual Studio 2005 x86 tools.
Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86

Copyright (C) Microsoft Corporation. All rights reserved.

hello.c
Setting environment for using Microsoft Visual Studio 2005 x64 tools.
Microsoft (R) Incremental Linker Version 8.00.50727.762
Copyright (C) Microsoft Corporation. All rights reserved.

C:\Project>hello.exe
Hello world!

Here: http://msdn.microsoft.com/en-us/library/5wy54dk2(v=VS.80).aspx

You might want to look into the /Machine:X86 option to the 64-bit linker...

Why can't you use the 64-bit toolchain from VS2005 to produce 32-bit binaries?

Are you dumb? ... You're using the VS2010 IDE to run the _exact same_ linker they are having problems with. How does that solve anything?

You're wrong, you can easily output 32-bit binaries with a 64-bit toolchain with MSVC - even from within the Visual Studio IDE if you set up the environment from the command line and then type "devenv /useenv". However you need to be running a 64-bit Windows which presumably the Firefox buildbots aren't.