Junio C Hamano <gitster@xxxxxxxxx> writes:
> So I would not mind if we define the semantics of "--filters" as
> such (as long as it is clearly documented, of course). AFAICS, the
> batch interface does not call filter_object() for non-blobs, and by
> returning successfully without doing anything special for a symbolic
> link from filter_object() automatically gives us the "by default
> return as-is, but give processed output only for regular file blobs"
> semantics to the batch mode.
>
> But for a non-batch mode, it feels somewhat funny to be giving the
> as-is output without saying anything to symbolic links; we can argue
> that it is being consistent with what we do in the batch mode,
> though.
In other words, instead of trying to be consistent by erroring out
in non-regular blob case, I think the attached change on top would
make more sense, by consistently passing the object contents as-is
for all "not filtered" cases, whether it is run from the batch mode
or from the command line.
builtin/cat-file.c | 7 +------
1 file changed, 1 insertion(+), 6 deletions(-)
diff --git a/builtin/cat-file.c b/builtin/cat-file.c
index f8a3a08..99cb525 100644
--- a/builtin/cat-file.c
+++ b/builtin/cat-file.c
@@ -33,12 +33,7 @@ static int filter_object(const char *path, unsigned mode,
if (!*buf)
return error(_("cannot read object %s '%s'"),
sha1_to_hex(sha1), path);
- if (type != OBJ_BLOB) {
- free(*buf);
- return error(_("blob expected for %s '%s'"),
- sha1_to_hex(sha1), path);
- }
- if (S_ISREG(mode)) {
+ if ((type == OBJ_BLOB) && S_ISREG(mode)) {
struct strbuf strbuf = STRBUF_INIT;
if (convert_to_working_tree(path, *buf, *size, &strbuf)) {
free(*buf);