threads / patch / 41586

patchbundle doc: 'verify' is not about verifying the bundle

Subject: [PATCH] bundle doc: 'verify' is not about verifying the bundle

## tl;dr

4 messages between Mar 1, 2016 and Mar 2, 2016. Diffs are folded; open one to read it.

replies: 3people: 3as markdown or json

Junio C Hamano· Mar 1, 2016, 21:36 UTC · lore

Even though the command does read the bundle header and checks to see if it looks reasonable, the thin-pack data stream that follows the header in the bundle file is not checked. More importantly, because the thin-pack data does not have a trailing checksum like on-disk packfiles do, there isn't much "verification" the command can do without unpacking the objects from the stream even if it wanted to.

The documentation gives an incorrect impression that the thin-pack data contained in the bundle is validated, but the command is to validate that the receiving repository is ready to accept the bundle, not to check the validity of a bundle file. Rephrase the paragraph to clarify this.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 Documentation/git-bundle.txt | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
Show changes to Documentation/git-bundle.txt +4 −5
diff --git a/Documentation/git-bundle.txt b/Documentation/git-bundle.txt
index 3a8120c..c0113a8 100644
--- a/Documentation/git-bundle.txt
+++ b/Documentation/git-bundle.txt
@@ -38,11 +38,10 @@ create <file>::
 	'git-rev-list-args' arguments to define the bundle contents.
 
 verify <file>::
-	Used to check that a bundle file is valid and will apply
-	cleanly to the current repository.  This includes checks on the
-	bundle format itself as well as checking that the prerequisite
-	commits exist and are fully linked in the current repository.
-	'git bundle' prints a list of missing commits, if any, and exits
+	Verifies that the given 'file' has a valid-looking bundle
+	header, and that your repository has all prerequisite
+	objects necessary to unpack the file as a bundle.  The
+	command prints a list of missing commits, if any, and exits
 	with a non-zero status.
 
 list-heads <file>::

-- 
2.8.0-rc0-114-g0b3e5e5
Philip Oakley· Mar 1, 2016, 22:49 UTC · re: Junio C Hamano · lore

Re: [PATCH] bundle doc: 'verify' is not about verifying the bundle

From: "Junio C Hamano" <gitster@pobox.com>
Show 14 quoted lines
> Even though the command does read the bundle header and checks to
> see if it looks reasonable, the thin-pack data stream that follows
> the header in the bundle file is not checked.  More importantly,
> because the thin-pack data does not have a trailing checksum like
> on-disk packfiles do, there isn't much "verification" the command
> can do without unpacking the objects from the stream even if it
> wanted to.
> 
> The documentation gives an incorrect impression that the thin-pack
> data contained in the bundle is validated, but the command is to
> validate that the receiving repository is ready to accept the
> bundle, not to check the validity of a bundle file.  Rephrase the
> paragraph to clarify this.
> 
This looks good to me.

I was actually looking at this over the weekend with respect to a back- burner issue about indicating which ref is HEAD within the bundle. (thread variously $gmane/265001 and it's prior links, though now I'm minded to add \0HEAD after the correct ref; e.g. '<sha1> refs/heads/<ref>\0HEAD')

Philip
Show 29 quoted lines
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> ---
> Documentation/git-bundle.txt | 9 ++++-----
> 1 file changed, 4 insertions(+), 5 deletions(-)
> 
> diff --git a/Documentation/git-bundle.txt b/Documentation/git-bundle.txt
> index 3a8120c..c0113a8 100644
> --- a/Documentation/git-bundle.txt
> +++ b/Documentation/git-bundle.txt
> @@ -38,11 +38,10 @@ create <file>::
>  'git-rev-list-args' arguments to define the bundle contents.
> 
> verify <file>::
> - Used to check that a bundle file is valid and will apply
> - cleanly to the current repository.  This includes checks on the
> - bundle format itself as well as checking that the prerequisite
> - commits exist and are fully linked in the current repository.
> - 'git bundle' prints a list of missing commits, if any, and exits
> + Verifies that the given 'file' has a valid-looking bundle
> + header, and that your repository has all prerequisite
> + objects necessary to unpack the file as a bundle.  The
> + command prints a list of missing commits, if any, and exits
>  with a non-zero status.
> 
> list-heads <file>::
> -- 
> 2.8.0-rc0-114-g0b3e5e5
> 
> --
Jeff King· Mar 2, 2016, 09:22 UTC · re: Junio C Hamano · lore

Re: [PATCH] bundle doc: 'verify' is not about verifying the bundle

On Tue, Mar 01, 2016 at 01:36:05PM -0800, Junio C Hamano wrote:
Show 7 quoted lines
> Even though the command does read the bundle header and checks to
> see if it looks reasonable, the thin-pack data stream that follows
> the header in the bundle file is not checked.  More importantly,
> because the thin-pack data does not have a trailing checksum like
> on-disk packfiles do, there isn't much "verification" the command
> can do without unpacking the objects from the stream even if it
> wanted to.
Are you sure about that trailing checksum thing?
  $ git tag foo $(echo foo | git hash-object -w --stdin)
  $ git bundle create foo.bundle foo
  $ perl -ne '
      BEGIN { binmode STDIN; binmode STDOUT }
      $print = 1 if /^PACK/;
      print if $print;
    ' <foo.bundle >foo.pack
  $ tail -c 20 foo.pack | xxd -p -c 20
  aa4e77be3695e8003bf4fbd2139b3fe0a9a22c8f
  $ git index-pack --stdin <foo.pack
  aa4e77be3695e8003bf4fbd2139b3fe0a9a22c8f

So I think we could pretty easily check for bit errors in the packfile (we can't in the ref data; including a separate sha1 there might be a nice improvement).

But...
Show 5 quoted lines
> The documentation gives an incorrect impression that the thin-pack
> data contained in the bundle is validated, but the command is to
> validate that the receiving repository is ready to accept the
> bundle, not to check the validity of a bundle file.  Rephrase the
> paragraph to clarify this.

I still think this is a good change, though. "verify" is doing a separate thing here than asking "did I get bit errors". We might _also_ want that check, but it is probably less interesting than "does the bundle have all of the objects it claims to". And you can only find that by expanding the objects, which would _also_ notice bit errors.

And the simplest way to do that is "git bundle apply". :)
-Peff
Junio C Hamano· Mar 2, 2016, 18:18 UTC · re: Jeff King · lore

Re: [PATCH] bundle doc: 'verify' is not about verifying the bundle

Jeff King <peff@peff.net> writes:
Show 11 quoted lines
> On Tue, Mar 01, 2016 at 01:36:05PM -0800, Junio C Hamano wrote:
>
>> Even though the command does read the bundle header and checks to
>> see if it looks reasonable, the thin-pack data stream that follows
>> the header in the bundle file is not checked.  More importantly,
>> because the thin-pack data does not have a trailing checksum like
>> on-disk packfiles do, there isn't much "verification" the command
>> can do without unpacking the objects from the stream even if it
>> wanted to.
>
> Are you sure about that trailing checksum thing?

No. I misread the fact that we do not say the final csum on the command output from pack-objects in the --stdout case. We do call the sha1close() to append the csum at the end of the pack stream; we just do not make it available to the caller who is driving the pack-object procedure.

← back to recent threads