On Tue, May 23, 2006 at 09:55:46PM +0200, Michael Kerrisk wrote:
> Justin,
>
> > Included is a manpage documenting the offsetof() macro; please
> > consider including it.
>
> Excellent idea! Thanks for the page. It will go in 2.34.
It was my idea .. I guess I found it in undocumented.3.
> > /* Output is compiler-dependant */
>
> Please put your pages through a spell checker. (A fault that
> I'm also sometimes guilty of...)
I know; I saw this last night. I'm building a spell check patch for
you. manpages are actually quite good at spelling :)
BTW:
$ grep -xi ^dependant /usr/share/dict/words
dependant
> > printf("%d\n", offsetof(struct a, c));
> > exit(EXIT_SUCCESS);
> > }
> > .fi
> > .SH NOTES
> > \fBoffsetof\fP can be implemented as:
> > .sp
> > .nf
> > \fB#define offsetof(\fItype\fP, \fPmember\fP) \\
> > \fB&((type *)0)->\fImember\fP - (char *)((\fPtype\fP *)0)
> > .fi
>
> I'm inclined to leave the above out (it could vary on some systems).
> But anyway, why do you write the above when glibc says:
>
> #define offsetof(TYPE, MEMBER) ((size_t) &((TYPE *)0)->MEMBER)
Actually, I have:
#define offsetof(TYPE, MEMBER) __builtin_offsetof (TYPE, MEMBER)
And in gcc-3.4 it is even uglier:
/* The cast to "char &" below avoids problems with user-defined
"operator &", which can appear in a POD type. */
#define offsetof(TYPE, MEMBER) \
(__offsetof__ (reinterpret_cast <size_t> \
(&reinterpret_cast <const volatile char &> \
(static_cast<TYPE *> (0)->MEMBER))))
But I don't know why it would be necessary to subtract the casted zero
:) At one point I had quite an amusing conversation about this.
I wrote it myself so I wasn't copying libc (although my source was gcc
anyway) and to make sure I still understood it. Do you have a glibc
source, too?
Justin
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
