it is possible to prevent implicit conversation in non-ctor-functions?

[pheres]

Hi,

consider a function taking a unsigned for example:

void func(unsigned _u)
{}

int main()
{
     func(-200); // gives garbage inside func
}

I know explicit decleration, which unfortunately only works for ctors using one argument.
Is there some else possiblitity besides configuring the compiler with something like "treat warnings as errors"?
And what is the reason explicit is not defined for ctors with more than one argument or other functions?

Thank You!

[cpjust]

Why wouldn't you always compile with the highest warning levels? Compiler warnings are your friend; if you ignore them, then whatever follows is your own fault.
But I do think C++ should allow the explicit keyword in every function, not just constructors. I hope they add that feature soon.

[pheres]

I do compile with all warnings I can get. For now I'm working on some library to give away and I want to help the users as much as I can to not run into slips.

[Elysia]

With all due respect, if they pass a signed integer into one where unsigned is expected, they're bound to run into problems. And it might also help to specify actual type, such as unsigned int.
From a language standpoint, unless you just typeid, there's no way to determine if someone has passed a correct argument and, indeed, restrict it.
And there's an option in most compilers that treats warnings as errors and refuses to compile, but I don't find it too big an issue for C++. I use it all the time with C, but not C++...

[cpjust]

And there's an option in most compilers that treats warnings as errors and refuses to compile, but I don't find it too big an issue for C++. I use it all the time with C, but not C++...

I use Warning Level 4 & Treat warnings as Errors all the time.
If you don't Treat Warnings as Errors, you'll get warnings the first time you compile, then if you click build again it doesn't report those warnings anymore until you do a Rebuild All.
Plus it forces lazy developers on your team to fix ALL warnings before they can create the executable.

[pheres]

What brings me to a releated question:
Do you check your build scripts/make files/solutions/whatever (with warning setup inside) into the VCS?

I do, but some guys beliefs into other compilers and ignores them.

[Elysia]

I always heed warnings unless I'm messing around with some non-serious code (like testing) and can probably leave unused variables in and such. That's also the primary reason I don't treat warnings like errors. If I don't need to fix them, then I don't.
But when doing serious coding, I always pay attention to them and fix them, as opposed to ignoring them.

[cpjust]

What brings me to a releated question:
Do you check your build scripts/make files/solutions/whatever (with warning setup inside) into the VCS?

I do, but some guys beliefs into other compilers and ignores them.

Yes, at my last company they created a /src directory for all the .h & .cpp files, then created a /x86-vc6 directory for VC6 Project & Workspace files, /x86-vc8 for VC++ 2005 files, /unix directory for makefiles...

My new company throws everything into one directory which is quite annoying.
All that gets checked into CVS, Source Safe, Perforce or whatever.

[CornedBee]

I check my jamfiles in.

You cannot prevent the compiler from performing the built-in implicit conversions. It's an important criticism of C and C++.

[pheres]

What brings me to a releated question:

If it's not tolerated inside the C++ forum, maybe it can be split off and moved to the tech board.

What build system do you like most and for what main reason? Has anyone experiences with SCONS?

[laserlight]

Please ask the build tool question in the Tech board. It is more general rather than C++ specific.

[robwhit]

And it might also help to specify actual type, such as unsigned int.

unsigned is identical to unsigned int. the int is implied.

[Elysia]

Perhaps, but I always find it better to be explicit, just like return int from main.