All of lore.kernel.org
 help / color / mirror / Atom feed
From: Derrick Stolee <derrickstolee@github.com>
To: Han Xin <hanxin.hx@bytedance.com>
Cc: git@vger.kernel.org, xingxin.xx@bytedance.com,
	jonathantanmy@google.com, Junio C Hamano <gitster@pobox.com>
Subject: Re: [External] Re: [PATCH v1] negotiator/default.c: avoid stack overflow
Date: Tue, 25 Apr 2023 09:34:38 -0400	[thread overview]
Message-ID: <4ddb4030-0789-5ed4-448d-96f6afb48334@github.com> (raw)
In-Reply-To: <CAKgqsWUEnbmhLL3p9+_P4yH_=A+hz+bBPqmfb8FyRUeW-u7_gw@mail.gmail.com>

On 4/24/2023 11:02 PM, Han Xin wrote:
> On Mon, Apr 24, 2023 at 10:44 PM Derrick Stolee
> <derrickstolee@github.com> wrote:

>>> @@ -70,15 +76,17 @@ static void mark_common(struct negotiation_state *ns, struct commit *commit,
>>>                               ns->non_common_revs--;
>>>                       if (!o->parsed && !dont_parse)
>>>                               if (repo_parse_commit(the_repository, commit))
>>> -                                     return;
>>> +                                     continue;
>>>
>>> +                     ancestors_only = 0;
>>
>> This caught me off guard, but this flag essentially says "should
>> I mark the first commit as common or not?". It would probably be
>> clearer if this was done before the loop, and then was ignored
>> within the loop, setting the flag on each parent in this loop:
>>
>>>                       for (parents = commit->parents;
>>>                                       parents;
>>>                                       parents = parents->next)
>>> -                             mark_common(ns, parents->item, 0,
>>> -                                         dont_parse);
>>> +                             prio_queue_put(&queue, parents->item);
>>
> 
> I'll think about how to optimize this again.
> 
> ancestors_only is used multiple times in the original logic:
> 1.
>               if (!ancestors_only)
>                      o->flags |= COMMON;
> 2.
>              if (!(o->flags & SEEN))
>                      rev_list_push(ns, commit, SEEN);
>              else {
>                      struct commit_list *parents;
> 
>                      if (!ancestors_only && !(o->flags & POPPED))
>                              ns->non_common_revs--;

Good point. Thanks for checking.
 
> Should we use this ?
> 
>              if (!ancestors_only) {
>                     commit->object.flags |= COMMON;
> 
>                     if ((commit->object.flags & SEEN) &&
> !(commit->object.flags & POPPED))
>                              ns->non_common_revs--;
>              }

This seems correct, although your email seems to have done a strange
line wrap that I'm sure you'll fix in the actual patch.

> and
> 
>                    for (parents = commit->parents;
>                              parents;
>                              parents = parents->next) {
>                              if (parents->item->object.flags & COMMON)
>                                       continue;
> 
>                             parents->item->object.flags |= COMMON;

Thanks, this part avoids duplicate additions to the queue.

>                             if ((parents->item->object.flags & SEEN)
>                                      && !(parents->item->object.flags & POPPED))
>                                       ns->non_common_revs--;

And this matches the non_common_revs part.

If you want this code to be a little cleaner, you could add

	struct commit *p = parents->item;

at the start of the loop and then s/parents->item/p/ for the rest
of the uses in the loop.

>                             prio_queue_put(&queue, parents->item);
>                    }

>> In addition, the mark_common() method there seems to have a few
>> problems:
...
>> Consider fixing these issues while you are here.
>>
> 
> Make sense.

I'm looking forward to your v2!

Thanks,
-Stolee

  reply	other threads:[~2023-04-25 13:35 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-04-24  2:23 [PATCH v1] negotiator/default.c: avoid stack overflow Han Xin
2023-04-24 14:44 ` Derrick Stolee
2023-04-25  3:02   ` [External] " Han Xin
2023-04-25 13:34     ` Derrick Stolee [this message]
2023-04-26  4:05 ` [PATCH v2 0/2] negotiator/default: " Han Xin
2023-04-26  4:05   ` [PATCH v2 1/2] " Han Xin
2023-04-26 11:13     ` Derrick Stolee
2023-04-26 11:40       ` [External] " Han Xin
2023-04-26  4:05   ` [PATCH v2 2/2] negotiator/skipping: fix some problems in mark_common() Han Xin
2023-04-26 11:08     ` Derrick Stolee
2023-04-26 11:55       ` [External] " Han Xin
2023-04-26 13:15   ` [PATCH v2 0/2] negotiator/default: avoid stack overflow Han Xin
2023-04-26 13:15     ` [PATCH v3 1/2] " Han Xin
2023-04-26 17:14       ` Junio C Hamano
2023-04-26 17:30         ` Derrick Stolee
2023-04-26 17:38           ` Junio C Hamano
2023-04-26 13:15     ` [PATCH v3 2/2] negotiator/skipping: fix some problems in mark_common() Han Xin
2023-05-01 22:11     ` [PATCH v2 0/2] negotiator/default: avoid stack overflow Junio C Hamano
2023-05-02  1:49       ` Derrick Stolee
2023-05-02 15:51         ` Junio C Hamano

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4ddb4030-0789-5ed4-448d-96f6afb48334@github.com \
    --to=derrickstolee@github.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    --cc=hanxin.hx@bytedance.com \
    --cc=jonathantanmy@google.com \
    --cc=xingxin.xx@bytedance.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.