The 02 November 2007 build (4592) is out.

[TDragon]

Nope, this is a bug in certain versions of windres. Wipe your 4.2.1 installation directory and reinstall with MinGW's binutils-2.17.50-20060824-1 package.

[doomleika]

However, I mean when you "stepped into the scope" the watch still displays wrong number. i.e.when stepped on function "int lol()" call, it's wrong number, when stepped on ++k, it still wrong number, so does k =34; std::cout << k;.........well but the "cout" still produce the correct result. That's why I suspect the watch is broken in some way.....

Nah, it's not a bug. ^^
++k is resolved after the debugger stopped there, that's the reason why it shows the false number.
If you would put a breakpoint the line after, it should show correct output.

You'll find the same behaviour in Microsofts Visual Studio, if I remember right.

again, thx for reply.

Well, it doesn't. If you put a breakpoint at int main() or {, in MSVC 2005 EE, the locale watch doen't even show the variable k util you stepped in the scope, it will display the random number before the variable initialization statment, and after the initialization, it works fine. If you force put on a watch manually it will say "CXX0069: Error: variable needs stack frame" before the scope.

If it is not a "bug", at least the IDE should stop user putting a breakpoint at main() or {.

[gd_on]

Thanks TDragon.
I had effectively a bad windres version (though it was from a binutils-2.17.50 but not the same date !!!).
Now it's OK. It compiles and the dll runs with C::B. Just one thing, my dll is 2.7 MO bigger than the one provided by killerbot. May be, he has stripped it.

gd_on

[ASJ]

Many, MANY thanks for fixing the "always rebuild all" bug! The problem is gone now and it is saving me a lot of time!
Thanks!

[pasgui]

Build for Ubuntu i386/amd64 can be found here

[stahta01]

Just one thing, my dll is 2.7 MO bigger than the one provided by killerbot. May be, he has stripped it.

Yes, he almost always strips the DLL.

Tim S