Re: minor fixes

[Rich Felker <dalias@aerifal.cx>]

On Sun, Sep 25, 2011 at 03:47:25PM +0200, Szabolcs Nagy wrote:
> attached a few lines of small changes
> 
> 
> * if the time.h fixme is because of the sigevent struct
> declaration, then it can go away as posix says
> "The tag sigevent shall be declared as naming an incomplete structure type,
> the contents of which are described in the <signal.h> header."
> 
> actually there is a much stronger statement as well:
> "Inclusion of the <time.h> header may make visible all symbols from the
> <signal.h> header."

OK, I guess this is what the comment was about. Actually I forgot what
it was there for. :-)

> * __syscall_ret: declaration should be consistent with the function
> definition (argument is signed long)
> 
> i think one of the (unsigned long) casts is unnecessary due to conversion
> rules but it probably makes things more clear

I want to keep the types from the function definition and fix the
prototype. They were basically chosen so that the "ugly"
(impl-defined) conversion (unsigned-to-signed) would always take place
inside the implementation of this function, and the safe
(signed-to-unsigned) conversion could take place implicitly in the
caller if necessary. It probably doesn't matter but I thought it
cleaner at the time.

> * sbrk: changed prototype here as well to be consistent with the standard
> 
> actually semantically ptrdiff_t is better (sbrk does p+arg, not p=arg)
> but the standard uses intptr_t
> 
> this change should not hurt as sbrk should not be used anyway..
> (i guess there is no architecture where intptr_t != ptrdiff_t)

I agree, this is the right fix.

> * __asctime: use new crash (i assume the *0=0 crash is deprecated now)

Good catch.

> * setpgid: pid_t return type is wrong (return value is an error code)

This one too.

Thanks!

Rich