Brian Downing schrieb:
> For some users (e.g. git blame), getting textual patch output is just
> extra work, as they can get all the information they need from the low-
> level diff structures. Allow for an alternate low-level emit function
> to be defined to allow bypassing the textual patch generation; set
> xemitconf_t's emit_func member to enable this.
>
> The (void (*)()) type is pretty ugly, but the alternative would be to
> include most of the private xdiff headers in xdiff.h to get the types
> required for the "proper" function prototype. Also, a (void *) won't
> work, as ANSI C doesn't allow a function pointer to be cast to an
> object pointer.
Could we move more code into the library code to avoid that ugliness?
AFAICS, compare_buffer() builds a struct patch with an array of
struct chunks, whose members are then fed one by one into either
blame_chunk() or handle_split(). Could we avoid the allocation
altogether by using a different interface?
E.g. have a callback like this:
static void handle_split_cb(long same, long p_next, long t_next,
void *data)
{
struct chunk_cb_data *d = data;
handle_split(d->sb, d->ent, d->tlno, d->plno, same,
d->parent, d->split);
d->plno = p_next;
d->tlno = t_next;
}
And use it like this:
struct chunk_cb_data d = {sb, ent, 0, 0, parent, split};
xpparam_t xpp;
xdemitconf_t xecfg;
xpp.flags = xdl_opts;
memset(&xecfg, 0, sizeof(xecfg));
xecfg.ctxlen = context;
xdi_diff_chunks(file_p, file_o, &xpp, &xecfg, handle_split_cb, &d);
handle_split(sb, ent, d.tlno, d.plno, ent->num_lines,
parent, split);
Makes sense?
René
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
