Re: [avr-libc-dev] [patch #6418] Yet another delay routines

[Carlos Lamas]

Hello,

I have not considered __builtin_constant_p(), I tend to forget all 
those built in GCC functions after several failed attempts to use some 
of them. I have been turning around your idea last days.

In my opinion the biggest weakness of my first proposed code is the 
obscure error message when the parameter passed is not a constant. It 
should help to present a clearer error message when the parameter is 
invalid. __builtin_constant_p() facilitates the testing but it does not 
allow to stop the compilation process by itself, and #error directive 
cannot be used in a macro expansion. The solution seems to use 
__builtin_constant_p() to conditionally compile the assembler directive 
'.error'. This way, when a non constant argument is entered the file 
compiles ok but the assembler stops showing errors. When combined with 
'.print' directive, and __FILE__ and __LINE__ symbols to clone an 
ordinary error message to help IDEs identifying the source of the error 
*.c file, the result looks pretty good. This method also allows to 
perform additional range checking in the delay parameter.

__builtin_constant_p() behaves slightly different depending on the 
optimization level selected. With -O0 'const' qualified variables are 
not recognized as constant values, and with optimization enabled 
'const' variables are seen as any other numerical constant. This 
introduces a little asymmetry.

In a couple of days or so I will complete testings and doxygen comment 
checking, to upload the new version to Savannah.

Best regards,

Carlos.


[...]

> > David Brown wrote:
> > > Carlos Lamas wrote:
> > > <snip>
> > >
> > > > Problem 2. The proposed solution is to check the constant parameter 
> > > > passed
> > > > against an inline assembler constraint in an empty dummy asm() 
> > > > instruction.
> > > > This also requires to transform the inline function in a macro. 
> > > > Passing a
> > > > variable as the argument to the macro generates an error a bit 
> > > > obfuscated.
> > >
> > > <snip>
> > >
> > > Hi Carlos,
> > >
> > > I don't know if you've considered it, but the 
> > > __builtin_constant_p(exp) function might be a better way to check 
> > > for constant parameters:
> > >
> > > http://gcc.gnu.org/onlinedocs/gcc-4.2.2/gcc/Other-Builtins.html
> > >
> > > [...]