On Sun, Sep 09, 2007 at 10:24:04PM +0000, Junio C Hamano wrote:
> Pierre Habouzit <madcoder@xxxxxxxxxx> writes:
>
> > +void strbuf_insert(struct strbuf *sb, size_t pos, const void *data, size_t len) {
> > + strbuf_grow(sb, len);
> > + if (pos >= sb->len) {
> > + sb->len = pos;
> > + } else {
> > + memmove(sb->buf + pos + len, sb->buf + pos, sb->len - pos);
> > + }
> > + memcpy(sb->buf + pos, data, len);
> > + strbuf_setlen(sb, sb->len + len);
> > +}
> > +
>
> What's the semantics of inserting way beyond len? If the buf is
> 5-byte "ABCDE" currently and you insert "FG" at position 7, do
> you get "ABCDE" + 2-byte undefined garbage + "FG" and result in
> 9-byte buffer? Or "FG" is appended after position 5 and you get
> "ABCDEFG" 7-byte buffer?
>
> I personally think we should OOPS such a caller, but I suspect
> you wanted the latter, in which case I think "sb->len = pos"
> assignment is done in the wrong direction, in other words, it
> should read "pos = sb->len" instead.
Yes, this was spotted by someone already. The semantics I attach is
that if pos is >= sb->len it appends (like it's commented in the .h).
Another patch I sent fixes that.
--
·O· Pierre Habouzit
··O madcoder@xxxxxxxxxx
OOO http://www.madism.org
Attachment:
pgp62Cu3PqSVg.pgp
Description: PGP signature
