git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 7/7] commit-graph.c: introduce '--[no-]check-oids'

From
SZEDER Gábor <szeder.dev@gmail.com>
Date
Apr 24, 2020, 10:59 UTC
Message-ID
<20200424105957.GB5925@szeder.dev>
In-Reply-To
<20200422233930.GB19100@syl.local>
On Wed, Apr 22, 2020 at 05:39:30PM -0600, Taylor Blau wrote:
Show 31 quoted lines
> > > diff --git a/Documentation/git-commit-graph.txt b/Documentation/git-commit-graph.txt
> > > index 46f7f7c573..91e8027b86 100644
> > > --- a/Documentation/git-commit-graph.txt
> > > +++ b/Documentation/git-commit-graph.txt
> > > @@ -82,6 +82,11 @@ tip with the previous tip.
> > >  Finally, if `--expire-time=<datetime>` is not specified, let `datetime`
> > >  be the current time. After writing the split commit-graph, delete all
> > >  unused commit-graph whose modified times are older than `datetime`.
> > > ++
> > > +The `--[no-]check-oids` option decides whether or not OIDs are required
> > > +to be commits. By default, `--check-oids` is implied, generating an
> > > +error on non-commit objects. If `--no-check-oids` is given, non-commits
> > > +are silently discarded.
> >
> > What happens with OIDs of tags, in particular with OIDs of tags that
> > can be peeled down to commit objects?  According to (my (too
> > pedantic?) interpretation of) the above description they will trigger
> > an error with '--check-oids' or will be ignored with
> > '--no-check-oids'.  The implementation, however, accepts those oids
> > and peels them down to commit objects; I think this is the right
> > behaviour.
> 
> That's right, and certainly merits a mention in the documentation. I've
> added that...
> 
> > What happens with OIDs that name non-existing objects?
> 
> ...these are silently discarded. I think that you could make a
> compelling argument in either direction on this one, but I'm slightly
> swayed towards "discard these, too", since '--no-check-oids' is
> literally saying "don't check these".

I don't want to argue either way, but I'd argue for making a conscious decision that is justified in the commit message and documented in the docs.

So, the option is '--stdin-commits' or '--input=stdin-commits', but it's not only about commits. Now, allowing OIDs of tags pointing to commits and peeling them makes obvious sense, because we want commits reachable from those tags included in the commit-graph file. Allowing OIDs of tags pointing to non-commits and silently ignoring them makes sense (to better support your 'git f-e-r ... | git c-g write ...' use case), though it's not that obvious (after all I managed to overlook it back then, that's why we are now here discussing these '--check-oids' patches).

But I'm not sure about silently ignoring OIDs pointing to non-existent objects, because those might be a sign of some issues in whatever is generating the list of objects to be fed to 'git c-g write'. E.g. there could be a race between 'git for-each-ref' listing OIDs and some other processes pruning them. Does this worth worrying about? Dunno... but at least let's think it through, and record in the commit message why we made that decision, whatever that decision might be.

> I guess that pushes us into the territory of whether or not "check" is
> the right verb. "verify"? 

Oh, I didn't think about this, but now that you mention it we have '--verify' in 'git rev-parse', 'git tag' and elsewhere, and we have 'verify-commit', 'verify-path' and 'verify-tag' commands. So '--verify-oids' might be more consistent. I kind of like the 'oids' suffix in the option name, though I don't know what else we might want to verify in this command in the future...

> "scrutinize" :)?
Huhh, erm, no ;)
> If you're otherwise satisfied with this series, here's the updated
> patch.

I haven't yet looked closely at the rest of the series... The documentation update in the updated patch below looks good to me, thanks.

Show 17 quoted lines
> -- >8 --
> 
> Subject: [PATCH] commit-graph.c: introduce '--[no-]check-oids'
> 
> When operating on a stream of commit OIDs on stdin, 'git commit-graph
> write' checks that each OID refers to an object that is indeed a commit.
> This is convenient to make sure that the given input is well-formed, but
> can sometimes be undesirable.
> 
> For example, server operators may wish to feed the full commit object
> IDs pointed to by refs that were updated during a push to 'git
> commit-graph write --input=stdin-commits', and silently discard any
> input that doesn't point at a commit. This can be done by combing the
> output of 'git for-each-ref' with '--format %(*objecttype)', but this
> requires opening up a potentially large number of objects.  Instead, it
> is more convenient to feed the updated object IDs to the commit-graph
> machinery, and let it throw out whatever remains.  to commits.

Either the bulk of a sentence is missing, or there is a stray(?) "to commits." at the end of this line.

Show 31 quoted lines
> Introduce '--[no-]check-oids' to make such a behavior possible. With
> '--check-oids' (the default behavior to retain backwards compatibility),
> 'git commit-graph write' will barf on a non-commit line in its input.
> With '--no-check-oids', such lines will be silently ignored, making the
> above possible by specifying this option.
> 
> No matter which is supplied, 'git commit-graph write' retains the
> behavior from the previous commit of rejecting non-OID inputs like
> "HEAD" and "refs/heads/foo" as before.
> 
> Signed-off-by: Taylor Blau <me@ttaylorr.com>
> ---
>  Documentation/git-commit-graph.txt |  6 ++++++
>  builtin/commit-graph.c             | 11 ++++++++---
>  t/t5318-commit-graph.sh            | 28 ++++++++++++++++++++++++++++
>  3 files changed, 42 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/git-commit-graph.txt b/Documentation/git-commit-graph.txt
> index 46f7f7c573..6bdbe42766 100644
> --- a/Documentation/git-commit-graph.txt
> +++ b/Documentation/git-commit-graph.txt
> @@ -82,6 +82,12 @@ tip with the previous tip.
>  Finally, if `--expire-time=<datetime>` is not specified, let `datetime`
>  be the current time. After writing the split commit-graph, delete all
>  unused commit-graph whose modified times are older than `datetime`.
> ++
> +The `--[no-]check-oids` option decides whether or not OIDs are required
> +to be commits. By default, `--check-oids` is implied, generating an
> +error on non-commit objects. If `--no-check-oids` is given, non-commits
> +and non-existent objects are silently discarded. In either case, tags
> +are peeled down to the object they reference.
Previous: Taylor BlauNext: Taylor Blau
Message 13 of 19 in “commit-graph: split strategies, '--[no-]check-oids'”
  1. 0/7 commit-graph: split strategies, '--[no-]check-oids'Taylor Blau, Apr 14, 2020
  2. 1/7 t/helper/test-read-graph.c: support commit-graph chainsTaylor Blau, Apr 14, 2020
  3. 2/7 builtin/commit-graph.c: support for '--split[=<strategy>]'Taylor Blau, Apr 14, 2020
  4. 3/7 builtin/commit-graph.c: introduce split strategy 'no-merge'Taylor Blau, Apr 14, 2020
  5. 4/7 builtin/commit-graph.c: introduce split strategy 'replace'Taylor Blau, Apr 14, 2020
  6. 5/7 oidset: introduce 'oidset_size'Taylor Blau, Apr 14, 2020
  7. 6/7 commit-graph.h: replace 'commit_hex' with 'commits'Taylor Blau, Apr 14, 2020
  8. 7/7 commit-graph.c: introduce '--[no-]check-oids'Taylor Blau, Apr 14, 2020
  9. Taylor BlauApr 15, 2020
  10. Taylor BlauApr 15, 2020
  11. SZEDER GáborApr 22, 2020
  12. Taylor BlauApr 22, 2020
  13. SZEDER GáborApr 24, 2020
  14. Taylor BlauMay 1, 2020
  15. Jeff KingMay 3, 2020
  16. Junio C HamanoMay 3, 2020
  17. Jeff KingMay 4, 2020
  18. Junio C HamanoMay 4, 2020
  19. Taylor BlauMay 4, 2020

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.