Re: full_read depends on incoming errno

[Eli Zaretskii]

> From: Andy Wingo <address@hidden>
> Cc: Eli Zaretskii <address@hidden>,  address@hidden,  address@hidden
> Date: Thu, 21 Feb 2013 17:47:43 +0100
> 
> On Thu 21 Feb 2013 17:29, Paolo Bonzini <address@hidden> writes:
> 
> > Il 20/02/2013 20:58, Andy Wingo ha scritto:
> >> 
> >>     if (full_read (fd, cookie, sizeof cookie) != sizeof cookie
> >>         || full_read (fd, SCM_BYTEVECTOR_CONTENTS (bv),
> >>                       SCM_BYTEVECTOR_LENGTH (bv)) != SCM_BYTEVECTOR_LENGTH 
> >> (bv))
> >>       {
> >>         int errno_save = errno;
> >>         (void) close (fd);
> >>         errno = errno_save;
> >>         if (errno)
> >>           SCM_SYSERROR;
> >>         scm_misc_error (FUNC_NAME, "file truncated while reading", 
> >> SCM_EOL);
> >>       }
> >
> > Doesn't it need to set errno to zero before calling full_read?
> 
> Not sure!  If full_read requires that errno be 0 going in, then yes, we
> should.  Is this a requirement?

AFAIK, you cannot rely on tests that compare errno with zero, unless
you reset it to zero before the operation that might set it.

> Otherwise, full_read sets errno to 0 if it returns a length less than
> the length we asked for

Where does it do that?  I don't think I see this.  Maybe I'm blind.