Re: Adding an option to avoid SIGPIPE for all file descriptors

[christos%zoulas.com@localhost (Christos Zoulas)]

On Jan 24,  5:19pm, dholland-tech%netbsd.org@localhost (David Holland) wrote:
-- Subject: Re: Adding an option to avoid SIGPIPE for all file descriptors

| On Mon, Jan 23, 2012 at 10:43:38PM -0500, Christos Zoulas wrote:
|  > | On Mon, Jan 23, 2012 at 09:28:48PM -0500, Christos Zoulas wrote:
|  > |  > [stuff]
|  > | 
|  > | One thing jumps out at me: if it's a file-level flag, shouldn't it be
|  > | sufficient to use F_GETFL / F_SETFL to manipulate it, rather than
|  > | adding new fcntls?
|  > 
|  > I thought about that, and we can definitely do this, but providing the
|  > fcntl's too gives us compatibility with the MacOSX api. I am open to
|  > whatever we decide.
| 
| I dunno. I'm generally opposed to multiplication of new fcntls or
| ioctls, but API compat does have some value... does anyone actually
| use that API?
| 
| (and I don't suppose we can think of a clever way to stuff the compat
| fcntls into libcompat or some such place so they don't bloat out the
| core system...)

The compat fcntls, are just 8 lines of code.

#define F_GETFOO X
#define F_SETFOO Y

case GETFOO:
        return (fp->f_flags & FFOO) != 0;

case SETFOO:
        fp->flags |= FFOO;
        break;


christos