On Sat, Feb 9, 2013 at 3:49 PM, Jay K <jay.kr...@cornell.edu> wrote:
> problems compiling 4.7, with Solaris cc and/or Solaris CC (C++)
>
>
> 1) ENUM_BITFIELD is not portable. I've reported this before.
>
>
> It should be more like:
>
>
> #ifdef __cplusplus
> #define ENUM_BITFIELD(TYPE, NAME, SIZE) enum TYPE NAME : SIZE
> #elif (GCC_VERSION > 2000)
> #define ENUM_BITFIELD(TYPE, NAME, SIZE) __extension__ enum TYPE NAME : SIZE
> #else
> #define ENUM_BITFIELD(TYPE, NAME, SIZE) unsigned int NAME : SIZE
> #endif
Could you explain why the current way is not correct?
Which does:
#ifdef __cplusplus
#define ENUM_BITFIELD(TYPE) enum TYPE
#elif (GCC_VERSION > 2000)
#define ENUM_BITFIELD(TYPE) __extension__ enum TYPE
#else
#define ENUM_BITFIELD(TYPE) unsigned int
#endif
Which is no different from what you have above really.
>
>
> and then modify the uses to fit -- three parameters.
>
>
> It is likely that in 4.8 this is moot, as the C++ case will be the only one
> remaining.
>
>
> 2) given:
>
> int foo()
> {
> gcc_unreachable();
> }
>
>
> Solaris cc/CC gives a warning or maybe an error.
That is a bug with Oracle's compilers as nothing after gcc_unreachable
is reachable
> It should be:
>
>
> int foo(void)
> {
> gcc_unreachable();
> return 0;
> }
>
>
> This occurs a few times e.g. in i386.md.
>
>
> Despite me being uncertain between warnings and errors and cc and CC,
> these do definitely show up as problems.
> I've been building a few versions of gcc through the years from 4.2
> and up (yeah, yeah, I know it goes back much further) with a variety
> of host compilers, mostly a few versions of gcc/g++ but sometimes Solaris
> cc/CC,
> maybe others, and I've had to patch these repeatedly. Again, today, in 4.7.
>
>
> I do use -disable-bootstrap, to drastically cut build times.
Cut build times but depends on bugs in the compiler which compiles stage1.
> It is possible that contributes to "the problem", but I also think these
> are reasonable changes and -disable-bootstrap is really nice to use, it
> saves a lot of time and these aren't awful ancient non-optimizing host
> compilers.
Except some host compilers have bugs. That is the point of the
bootstrap rather than the non-optimizing host compilers. As show
above Oracle's compilers have bugs already.
Thanks,
Andrew Pinski
