conversion of the __inline__ and inline pragmas to RTEMS_INLINE_ROUTINE?

[Ralf Corsepius]

On 11/09/2011 04:21 PM, Cudmore, Alan P. (GSFC-5820) wrote:
> My code needs to have the __inline__ pragma. The inline will not compiler with the switches that I am required to use.
> Our coding standard dictates:
>       -Wall -Wstrict-prototypes -pedantic -ansi -Werror

Well, IMO, these coding standards are pretty questionable for several 
reasons:

a) All that -Werror does is to raise arbitrary warnings, which your 
version of gcc believes are "questionable pieces of code" to errors.
- This doesn't add any new information to the warnings the other flags 
already trigger.
- The gcc raises warnings for contain "false warnings" and "stylistic 
warning". The more aggressive flags you are using, the more such 
"neglible warnings" you will see.
- The warnings gcc raises are gcc-version dependent.
I.e. even if you manage to get things compiled without any warning with 
one version of gcc, different versions of GCC will raise different 
warnings and let your compilation fail due to -Werror.

b) -ansi will turn off prototypes of functions in newlib, RTEMS 
currently relies upon (This is the c89/c99 compatibility issue in 
newlib), I just mentioned in another mail. I.e. you are likely to be 
facing "implicit decls" due to -ansi, and are not unlikely to be facing 
miscompilation.

c) -pedantic (and -Wextra) lets GCC trigger warnings, it itself doesn't 
consider to be "fail safe" => it will trigger further "false warnings".

d) -Wstrict-prototypes is good thing and prevents a lot of issues.

Ralf