git@vger.kernel.org list mirror (unofficial, one of many)
 help / color / Atom feed
From: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
To: Derrick Stolee <stolee@gmail.com>
Cc: Derrick Stolee via GitGitGadget <gitgitgadget@gmail.com>,
	git@vger.kernel.org, sandals@crustytoothpaste.net,
	Junio C Hamano <gitster@pobox.com>, Jeff King <peff@peff.net>
Subject: Re: [PATCH v3 0/6] Create commit-graph file format v2
Date: Fri, 03 May 2019 15:41:36 +0200
Message-ID: <87bm0jirjj.fsf@evledraar.gmail.com> (raw)
In-Reply-To: <04ac97c5-ab2b-2b23-4c84-8ed8aab0e2f5@gmail.com>


On Fri, May 03 2019, Derrick Stolee wrote:

> On 5/2/2019 2:02 PM, Ævar Arnfjörð Bjarmason wrote:
>>
>> But those are separate from any back-compat concerns, which is what I
>> think makes sense to focus on now.
>
> Thinking more on this topic, I think I have a way to satisfy _all_ of
> your concerns by simplifying the plan for incremental commit-graph files.
>
> My initial plan was to have the "commit-graph" file always be the "tip"
> file, and it would point to some number of "graph-{hash}.graph" files.
> Then, we would have some set of strategies to decide when we create a new
> .graph file or when we compact the files down into the "commit-graph"
> file.
>
> This has several issues regarding race conditions that I had not yet
> resolved (and maybe would always have problems).
>
> It would be much simpler to restrict the model. Your idea of changing
> the file name is the inspiration here.

I still have some questions about SHA-256, how we'll eventually change
the format of these new files etc, but those can wait...

> * The "commit-graph" file is the base commit graph. It is always
>   closed under reachability (if a commit exists in this file, then
>   its parents are also in this file). We will also consider this to
>   be "commit-graph-0".
>
> * A commit-graph-<N> exists, then we check for the existence of
>   commit-graph-<N+1>. This file can contain commits whose parents
>   are in any smaller file.
>
> I think this resolves the issue of back-compat without updating
> the file format:
>
> 1. Old clients will never look at commit-graph-N, so they will
>    never complain about an "incomplete" file.
>
> 2. If we always open a read handle as we move up the list, then
>    a "merge and collapse" write to commit-graph-N will not
>    interrupt an existing process reading that file.
>
> I'll start hacking on this model.

Just on this, consider storing them in
.git/objects/info/commit-graphs/commit-graph-<THIS-FILE'S-CHECKSUM-SHA1>,
because:

1) We can stat() the "commit-graphs" directory to see if there's any
   new/deleted ones (dir mtime changed), similar to what we do for the
   untracked cache, and can (but I don't think we do...) do for packs/*
   and objects/??/.

   As opposed to ".git/objects/info" itself which e.g. has the "packs",
   "alternates" etc. files (can still do it, but more false positives)

2) If these things are "chained" supersets of one another anyway we have
   a big caveat that we don't have with repacking the *.pack
   files.

   Those you can do in any order, as long as you write a new one it
   doesn't matter if the data is already elsewhere stored. Just repack
   however many N at a time, then unlink() all the old ones later.

   If you have commit-graph-<N>, commit-graph-<N+1> you can only munge
   them in that sequence, and you get to do a much more careful fsync()
   dance around with both the directory entry and the file itself, which
   has a much higher chance of breaking things.

   I.e. the client might see a change to "N" before "N+1", and now we're
   screwed. Dealing with this is a giant but avoidable PITA. See
   https://public-inbox.org/git/20180117184828.31816-1-hch@lst.de/ and
   http://blog.httrack.com/blog/2013/11/15/everything-you-always-wanted-to-know-about-fsync/

   If instead you name the files commit-graph-<MY-OWN-SHA>, and add a
   new (v1 backwards-compatible) chunk for "and the next file's SHA-1
   is..."  we can more aggressively repack these.

   Worst case the directory entries will sync in the wrong order and the
   chain will be broken, but as long as we gracefully handle the chain
   abruptly ending that's OK.

   You can also do things like rewrite the middle "[]" of a "N, [N+1,
   N+2], N+3" sequence to have a single file for what was previously
   within those brackets (again, like "repack"). You'd just need to drop
   in a new "N" that points to the replacement for "[N+1, N+2]", but
   *don't* need to touch "N+3".

   Whereas if they're numbered you need to also move "N+3", and you're
   back to juggling both fsync() on files and directory entries.

  reply index

Thread overview: 89+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-01-23 21:59 [PATCH " Derrick Stolee via GitGitGadget
2019-01-23 21:59 ` [PATCH 1/6] commit-graph: return with errors during write Derrick Stolee via GitGitGadget
2019-01-23 21:59 ` [PATCH 2/6] commit-graph: collapse parameters into flags Derrick Stolee via GitGitGadget
2019-01-23 21:59 ` [PATCH 3/6] commit-graph: create new version flags Derrick Stolee via GitGitGadget
2019-01-23 21:59 ` [PATCH 4/6] commit-graph: add --version=<n> option Derrick Stolee via GitGitGadget
2019-01-24  9:31   ` Ævar Arnfjörð Bjarmason
2019-01-23 21:59 ` [PATCH 5/6] commit-graph: implement file format version 2 Derrick Stolee via GitGitGadget
2019-01-23 23:56   ` Jonathan Tan
2019-01-24  9:40   ` Ævar Arnfjörð Bjarmason
2019-01-24 14:34     ` Derrick Stolee
2019-03-21  9:21   ` Ævar Arnfjörð Bjarmason
2019-01-23 21:59 ` [PATCH 6/6] commit-graph: test verifying a corrupt v2 header Derrick Stolee via GitGitGadget
2019-01-23 23:59   ` Jonathan Tan
2019-01-24 23:05 ` [PATCH 0/6] Create commit-graph file format v2 Junio C Hamano
2019-01-24 23:39 ` Junio C Hamano
2019-01-25 13:54   ` Derrick Stolee
2019-04-24 19:58 ` [PATCH v2 0/5] " Derrick Stolee via GitGitGadget
2019-04-24 19:58   ` [PATCH v2 1/5] commit-graph: return with errors during write Derrick Stolee via GitGitGadget
2019-04-24 19:58   ` [PATCH v2 2/5] commit-graph: collapse parameters into flags Derrick Stolee via GitGitGadget
2019-04-25  5:21     ` Junio C Hamano
2019-04-24 19:58   ` [PATCH v2 3/5] commit-graph: create new version flags Derrick Stolee via GitGitGadget
2019-04-25  5:29     ` Junio C Hamano
2019-04-25 11:09       ` Derrick Stolee
2019-04-25 21:31     ` Ævar Arnfjörð Bjarmason
2019-04-26  2:20       ` Junio C Hamano
2019-04-24 19:58   ` [PATCH v2 4/5] commit-graph: add --version=<n> option Derrick Stolee via GitGitGadget
2019-04-24 19:58   ` [PATCH v2 5/5] commit-graph: implement file format version 2 Derrick Stolee via GitGitGadget
2019-04-25 22:09   ` [PATCH v2 0/5] Create commit-graph file format v2 Ævar Arnfjörð Bjarmason
2019-04-26  2:28     ` Junio C Hamano
2019-04-26  8:33       ` Ævar Arnfjörð Bjarmason
2019-04-26 12:06         ` Derrick Stolee
2019-04-26 13:55           ` Ævar Arnfjörð Bjarmason
2019-04-27 12:57     ` Ævar Arnfjörð Bjarmason
2019-05-01 13:11   ` [PATCH v3 0/6] " Derrick Stolee via GitGitGadget
2019-05-01 13:11     ` [PATCH v3 1/6] commit-graph: return with errors during write Derrick Stolee via GitGitGadget
2019-05-01 14:46       ` Ævar Arnfjörð Bjarmason
2019-05-01 13:11     ` [PATCH v3 2/6] commit-graph: collapse parameters into flags Derrick Stolee via GitGitGadget
2019-05-01 13:11     ` [PATCH v3 3/6] commit-graph: create new version parameter Derrick Stolee via GitGitGadget
2019-05-01 13:11     ` [PATCH v3 4/6] commit-graph: add --version=<n> option Derrick Stolee via GitGitGadget
2019-05-01 13:11     ` [PATCH v3 5/6] commit-graph: implement file format version 2 Derrick Stolee via GitGitGadget
2019-05-01 19:12       ` Ævar Arnfjörð Bjarmason
2019-05-01 19:56         ` Derrick Stolee
2019-05-01 13:11     ` [PATCH v3 6/6] commit-graph: remove Future Work section Derrick Stolee via GitGitGadget
2019-05-01 14:58       ` Ævar Arnfjörð Bjarmason
2019-05-01 19:59         ` Derrick Stolee
2019-05-01 20:25     ` [PATCH v3 0/6] Create commit-graph file format v2 Ævar Arnfjörð Bjarmason
2019-05-02 13:26       ` Derrick Stolee
2019-05-02 18:02         ` Ævar Arnfjörð Bjarmason
2019-05-03 12:47           ` Derrick Stolee
2019-05-03 13:41             ` Ævar Arnfjörð Bjarmason [this message]
2019-05-06  8:27               ` Christian Couder
2019-05-06 13:47                 ` Derrick Stolee
2019-05-03 14:16             ` SZEDER Gábor
2019-05-03 15:11               ` Derrick Stolee
2019-05-09 14:22     ` [PATCH v4 00/11] Commit-graph write refactor (was: Create commit-graph file format v2) Derrick Stolee via GitGitGadget
2019-05-09 14:22       ` [PATCH v4 01/11] commit-graph: fix the_repository reference Derrick Stolee via GitGitGadget
2019-05-13  2:56         ` Junio C Hamano
2019-05-09 14:22       ` [PATCH v4 02/11] commit-graph: return with errors during write Derrick Stolee via GitGitGadget
2019-05-13  3:13         ` Junio C Hamano
2019-05-13 11:04           ` Derrick Stolee
2019-05-13 11:22             ` Derrick Stolee
2019-05-09 14:22       ` [PATCH v4 03/11] commit-graph: collapse parameters into flags Derrick Stolee via GitGitGadget
2019-05-13  3:44         ` Junio C Hamano
2019-05-13 11:07           ` Derrick Stolee
2019-05-09 14:22       ` [PATCH v4 04/11] commit-graph: remove Future Work section Derrick Stolee via GitGitGadget
2019-05-09 14:22       ` [PATCH v4 05/11] commit-graph: create write_commit_graph_context Derrick Stolee via GitGitGadget
2019-05-09 14:22       ` [PATCH v4 07/11] commit-graph: extract fill_oids_from_commit_hex() Derrick Stolee via GitGitGadget
2019-05-09 14:22       ` [PATCH v4 06/11] commit-graph: extract fill_oids_from_packs() Derrick Stolee via GitGitGadget
2019-05-13  5:05         ` Junio C Hamano
2019-05-09 14:22       ` [PATCH v4 08/11] commit-graph: extract fill_oids_from_all_packs() Derrick Stolee via GitGitGadget
2019-05-09 14:22       ` [PATCH v4 09/11] commit-graph: extract count_distinct_commits() Derrick Stolee via GitGitGadget
2019-05-09 14:22       ` [PATCH v4 10/11] commit-graph: extract copy_oids_to_commits() Derrick Stolee via GitGitGadget
2019-05-09 14:22       ` [PATCH v4 11/11] commit-graph: extract write_commit_graph_file() Derrick Stolee via GitGitGadget
2019-05-13  5:09         ` Junio C Hamano
2019-05-09 17:58       ` [PATCH v4 00/11] Commit-graph write refactor (was: Create commit-graph file format v2) Josh Steadmon
2019-06-12 13:29       ` [PATCH v5 " Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 01/11] commit-graph: fix the_repository reference Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 02/11] commit-graph: return with errors during write Derrick Stolee via GitGitGadget
2019-06-29 17:23           ` SZEDER Gábor
2019-07-01 12:19             ` Derrick Stolee
2019-06-12 13:29         ` [PATCH v5 03/11] commit-graph: collapse parameters into flags Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 04/11] commit-graph: remove Future Work section Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 05/11] commit-graph: create write_commit_graph_context Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 06/11] commit-graph: extract fill_oids_from_packs() Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 07/11] commit-graph: extract fill_oids_from_commit_hex() Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 08/11] commit-graph: extract fill_oids_from_all_packs() Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 09/11] commit-graph: extract count_distinct_commits() Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 10/11] commit-graph: extract copy_oids_to_commits() Derrick Stolee via GitGitGadget
2019-06-12 13:29         ` [PATCH v5 11/11] commit-graph: extract write_commit_graph_file() Derrick Stolee via GitGitGadget

Reply instructions:

You may reply publically 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

  List information: http://vger.kernel.org/majordomo-info.html

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

  git send-email \
    --in-reply-to=87bm0jirjj.fsf@evledraar.gmail.com \
    --to=avarab@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=gitgitgadget@gmail.com \
    --cc=gitster@pobox.com \
    --cc=peff@peff.net \
    --cc=sandals@crustytoothpaste.net \
    --cc=stolee@gmail.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

git@vger.kernel.org list mirror (unofficial, one of many)

Archives are clonable:
	git clone --mirror https://public-inbox.org/git
	git clone --mirror http://ou63pmih66umazou.onion/git
	git clone --mirror http://czquwvybam4bgbro.onion/git
	git clone --mirror http://hjrcffqmbrq6wope.onion/git

Example config snippet for mirrors

Newsgroups are available over NNTP:
	nntp://news.public-inbox.org/inbox.comp.version-control.git
	nntp://ou63pmih66umazou.onion/inbox.comp.version-control.git
	nntp://czquwvybam4bgbro.onion/inbox.comp.version-control.git
	nntp://hjrcffqmbrq6wope.onion/inbox.comp.version-control.git
	nntp://news.gmane.org/gmane.comp.version-control.git

 note: .onion URLs require Tor: https://www.torproject.org/

AGPL code for this site: git clone https://public-inbox.org/public-inbox.git