Re: [PATCH 09/15] http: release the memory of a http pack request as well

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 24, 2015, 17:48 UTC
Message-ID: <xmqqh9ta70dc.fsf@gitster.dls.corp.google.com>
In-Reply-To: <CAGZ79kZbe9d7wG+CJ0o93hDs3sFjkSp68pBHXpjOysbzgvjo3w@mail.gmail.com>

Stefan Beller <sbeller@google.com> writes:

Show 5 quoted lines

> Well there is hope, as `release_request` only touches
>     free(request->url);
>     free(request);
>
> and not the userData pointer.

OK.

> I am a bit puzzled what you're trying to hint at.

The caller does this:

        static void start_fetch_packed(struct transfer_request *request)
        {
                ...
                preq = new_http_pack_request(target, repo->url);
                ...
                preq->slot->callback_func = process_response;
                preq->slot->callback_data = request;
                request->slot = preq->slot;
                request->userData = preq;

                /* Try to get the request started, abort the request on error */
                request->state = RUN_FETCH_PACKED;
                if (!start_active_slot(preq->slot)) {
                        fprintf(stderr, "Unable to start GET request\n");
                        release_http_pack_request(preq);
                        repo->can_update_info_refs = 0;
                        release_request(request);
                }
        }

and start_active_slot() actually not just "starts" but calls curl_multi_perform() to do things, like calling process_response(), which has calls to release_{,http_pack_}request(). I didn't see those releases and the releases we see in the above (i.e. when !start) will not run at the same time (but I see it now ;-))

In short, not hinting at anything. I was genuinely having a hard time following the codeflow.