Re: [PATCH v7 1/4] sha1_file.c: support reading from a loose object of unknown type
- From
karthik nayak <karthik.188@gmail.com>
- Date
- Apr 7, 2015, 10:34 UTC
- Message-ID
- <5523B2A7.6060604@gmail.com>
- In-Reply-To
- <xmqqbnj2e4ah.fsf@gitster.dls.corp.google.com>
On 04/06/2015 01:27 AM, Junio C Hamano wrote:
Show 77 quoted lines
> karthik nayak <karthik.188@gmail.com> writes:
>
> > On 04/05/2015 01:16 PM, Junio C Hamano wrote:
> >
> >> If it semantically does not make sense to ask for the typename
> >> without asking for the type code, then we can and should make that
> >> as a new calling convention _all_ callers must follow.
> >>
> >> In other words, I think it is better to have
> >>
> >> if (oi->typename && !oi->typep)
> >> die("BUG");
> >>
> >> somewhere near the beginning of the callchain that takes oi; that
> >> will ensure all callers understand the rule.
> >
> > Yes! this is a better approach as it will enforce that typep must be
> > set when typename is set.
>
> Not so fast ;-)
>
> The key phrase in what I wrote above is "If it does not make sense",
> and I am not yet convinced if that is the case or not. If we are to
> change the calling convention for the callers, the reason why it
> does not make sense to ask only for typename but not typep must be
> explained in the log message.
>
> And if it turns out that the answer to that question is "it is valid
> to ask only for typename", then it would be wrong to force everybody
> who wants to learn the stringified typename to supply typep. And in
> such a case it might be better to do something like this instead (on
> top of your patch I am responding to):
>
> sha1_file.c | 12 ++++++++++++
> 1 file changed, 12 insertions(+)
>
> diff --git a/sha1_file.c b/sha1_file.c
> index f4055dd..26fbb7c 100644
> --- a/sha1_file.c
> +++ b/sha1_file.c
> @@ -2639,6 +2639,7 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
> struct cached_object *co;
> struct pack_entry e;
> int rtype;
> + enum object_type real_type;
> const unsigned char *real = lookup_replace_object_extended(sha1, flags);
>
> co = find_cached_object(real);
> @@ -2670,9 +2671,18 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
> return -1;
> }
>
> + /*
> + * packed_object_info() does not follow the delta chain to
> + * find out the real type, unless it is given oi->typep.
> + */
> + if (oi->typename && !oi->typep)
> + oi->typep = &real_type;
> +
> rtype = packed_object_info(e.p, e.offset, oi);
> if (rtype < 0) {
> mark_bad_packed_object(e.p, real);
> + if (oi->typep == &real_type)
> + oi->typep = NULL;
> return sha1_object_info_extended(real, oi, 0);
> } else if (in_delta_base_cache(e.p, e.offset)) {
> oi->whence = OI_DBCACHED;
> @@ -2686,6 +2696,8 @@ int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi,
> if (oi->typename)
> strbuf_addstr(oi->typename, typename(*oi->typep));
>
> + if (oi->typep == &real_type)
> + oi->typep = NULL;
> return 0;
> }
>
>Haha, If there is anything I love about code revisions its how you are subjected to so many different ways of thinking. I didn't think of what you said.
We could do this and support typename without typep being used. This could probably even eliminate the typep used by cat-file while getting the type of an object.
Will go through this again.
Thanks a lot.