threads / patch / 46474

patchsubmodule: correct error message for missing commits.

Subject: [PATCH] submodule: correct error message for missing commits.

## tl;dr

4 messages between Jul 26, 2017 and Jul 26, 2017. Diffs are folded; open one to read it.

replies: 3people: 2as markdown or json

Stefan Beller· Jul 26, 2017, 20:08 UTC · lore

When a submodule diff should be displayed we currently just add the submodule objects to the main object store and then e.g. walk the revision graph and create a summary for that submodule.

It is possible that we are missing the submodule either completely or partially, which we currently differentiate with different error messages depending on whether (1) the whole submodule object store is missing or (2) just the needed for this particular diff. (1) is reported as "not initialized", and (2) is reported as "commits not present".

If a submodule is deinit'ed its repository data is still around inside the superproject, such that the diff can still be produced. In that way the error message (1) is misleading as we can have a diff despite the submodule being not initialized.

Downgrade the error message (1) to be the same as (2) and just say the commits are not present, as that is the true reason why the diff cannot be shown.

Signed-off-by: Stefan Beller <sbeller@google.com>
---

I came across this error message in the series for the object store modularisation[1], when I was trying to replace 'add_submodule_odb' by a custom loaded object store from a submodule repo object, which got me thinking on the error message and the true cause for it.

While this could go in separately, I may carry it in that series, as there we'd come up with more error messages ("could not create submodule object store" as well as the "commits not present", maybe even "submodule not lookup failed")

Thanks, Stefan

[1] https://public-inbox.org/git/20170706202739.6056-1-sbeller@google.com/

 submodule.c                               | 2 +-
 t/t4059-diff-submodule-not-initialized.sh | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

Show changes to 2 files +2 −2

submodule.c, t/t4059-diff-submodule-not-initialized.sh

diff --git a/submodule.c b/submodule.c
index 6531c5d609..280c246477 100644
--- a/submodule.c
+++ b/submodule.c
@@ -567,7 +567,7 @@ static void show_submodule_header(FILE *f, const char *path,
 
 	if (add_submodule_odb(path)) {
 		if (!message)
-			message = "(not initialized)";
+			message = "(commits not present)";
 		goto output_header;
 	}
 
diff --git a/t/t4059-diff-submodule-not-initialized.sh b/t/t4059-diff-submodule-not-initialized.sh
index cd70fd5192..49bca7b48d 100755
--- a/t/t4059-diff-submodule-not-initialized.sh
+++ b/t/t4059-diff-submodule-not-initialized.sh
@@ -95,7 +95,7 @@ test_expect_success 'submodule not initialized in new clone' '
 	git clone . sm3 &&
 	git -C sm3 diff-tree -p --no-commit-id --submodule=log HEAD >actual &&
 	cat >expected <<-EOF &&
-	Submodule sm1 $smhead1...$smhead2 (not initialized)
+	Submodule sm1 $smhead1...$smhead2 (commits not present)
 	EOF
 	test_cmp expected actual
 '

-- 
2.14.0.rc0.3.g6c2e499285

Junio C Hamano· Jul 26, 2017, 20:30 UTC · re: Stefan Beller · lore

Re: [PATCH] submodule: correct error message for missing commits.

Stefan Beller <sbeller@google.com> writes:

Show 14 quoted lines

> When a submodule diff should be displayed we currently just add the
> submodule objects to the main object store and then e.g. walk the
> revision graph and create a summary for that submodule.
>
> It is possible that we are missing the submodule either completely or
> partially, which we currently differentiate with different error messages
> depending on whether (1) the whole submodule object store is missing or
> (2) just the needed for this particular diff. (1) is reported as
> "not initialized", and (2) is reported as "commits not present".
>
> If a submodule is deinit'ed its repository data is still around inside
> the superproject, such that the diff can still be produced. In that way
> the error message (1) is misleading as we can have a diff despite the
> submodule being not initialized.

This is confusing...

So are you saying that if we do "submodule init A && submodule update A" followed by "submodule deinit A", we _could_ show the difference for submodule A between two commits in the superproject, because we already have the necessary data for the submodule, but we _choose_ not to show it because the user told us explicitly that the submodule is not interesting?

That sounds like a very sensible and user-centric behaviour to me, and "not initialized" sounds like the right message to give in such a case (as opposed to "commits not present"---even the user told us they are not interesting, we may have them, so "not present" is not just incorrect but irrelevant because that is not the reason why we are not showing).

Or are you saying that even the user told us that the submodule is not interesting, if we had "init" it earlier even once, we show the difference and with a wrong label? Showing the difference sounds like a bug that is more severe than using a wrong label to me.

Puzzled.

Show 55 quoted lines

>
> Downgrade the error message (1) to be the same as (2) and just say
> the commits are not present, as that is the true reason why the diff
> cannot be shown.
>
> Signed-off-by: Stefan Beller <sbeller@google.com>
> ---
>
> I came across this error message in the series for the
> object store modularisation[1], when I was trying to replace
> 'add_submodule_odb' by a custom loaded object store from a
> submodule repo object, which got me thinking on the error
> message and the true cause for it.  
>
> While this could go in separately, I may carry it in that
> series, as there we'd come up with more error messages
> ("could not create submodule object store" as well as the
> "commits not present", maybe even "submodule not lookup failed")
>
> Thanks,
> Stefan
>
> [1] https://public-inbox.org/git/20170706202739.6056-1-sbeller@google.com/
>   
>
>  submodule.c                               | 2 +-
>  t/t4059-diff-submodule-not-initialized.sh | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/submodule.c b/submodule.c
> index 6531c5d609..280c246477 100644
> --- a/submodule.c
> +++ b/submodule.c
> @@ -567,7 +567,7 @@ static void show_submodule_header(FILE *f, const char *path,
>  
>  	if (add_submodule_odb(path)) {
>  		if (!message)
> -			message = "(not initialized)";
> +			message = "(commits not present)";
>  		goto output_header;
>  	}
>  
> diff --git a/t/t4059-diff-submodule-not-initialized.sh b/t/t4059-diff-submodule-not-initialized.sh
> index cd70fd5192..49bca7b48d 100755
> --- a/t/t4059-diff-submodule-not-initialized.sh
> +++ b/t/t4059-diff-submodule-not-initialized.sh
> @@ -95,7 +95,7 @@ test_expect_success 'submodule not initialized in new clone' '
>  	git clone . sm3 &&
>  	git -C sm3 diff-tree -p --no-commit-id --submodule=log HEAD >actual &&
>  	cat >expected <<-EOF &&
> -	Submodule sm1 $smhead1...$smhead2 (not initialized)
> +	Submodule sm1 $smhead1...$smhead2 (commits not present)
>  	EOF
>  	test_cmp expected actual
>  '

Stefan Beller· Jul 26, 2017, 20:56 UTC · re: Junio C Hamano · lore

Re: [PATCH] submodule: correct error message for missing commits.

On Wed, Jul 26, 2017 at 1:30 PM, Junio C Hamano <gitster@pobox.com> wrote:

Show 21 quoted lines

> Stefan Beller <sbeller@google.com> writes:
>
>> When a submodule diff should be displayed we currently just add the
>> submodule objects to the main object store and then e.g. walk the
>> revision graph and create a summary for that submodule.
>>
>> It is possible that we are missing the submodule either completely or
>> partially, which we currently differentiate with different error messages
>> depending on whether (1) the whole submodule object store is missing or
>> (2) just the needed for this particular diff. (1) is reported as
>> "not initialized", and (2) is reported as "commits not present".
>>
>> If a submodule is deinit'ed its repository data is still around inside
>> the superproject, such that the diff can still be produced. In that way
>> the error message (1) is misleading as we can have a diff despite the
>> submodule being not initialized.
>
> This is confusing...
>
> So are you saying that if we do "submodule init A && submodule
> update A" followed by "submodule deinit A",

  $ git clone https://gerrit.googlesource.com/gerrit
  $ git show --submodule=log bb798b00bb
  ...
  Submodule plugins/replication ... (not initialized)

  $ git submodule update --init
  $ # a good example of cross repo changes:
  $ git show --submodule=log bb798b00bb
  ...
  Submodule plugins/replication db4aecb2b8...98b7156cee:
  > Stop using WorkQueue#unregisterWorkQueue.
  < PushOne: Remove redundant string concatenation

  $ git submodule deinit -f --all
  $ git show --submodule=log bb798b00bb
  ...
  Submodule plugins/replication db4aecb2b8...98b7156cee:
  > Stop using WorkQueue#unregisterWorkQueue.
  < PushOne: Remove redundant string concatenation

  $ rm -rf .git/modules/*
  $ git show --submodule=log bb798b00bb
  ...
  Submodule plugins/replication ... (not initialized)

Show 5 quoted lines

> we _could_ show the
> difference for submodule A between two commits in the superproject,
> because we already have the necessary data for the submodule, but we
> _choose_ not to show it because the user told us explicitly that the
> submodule is not interesting?

We _do_ show the submodule as demonstrated by the code sample above if we possess the objects.

Hence the "not initialized" is not quite technically correct. (After deinit it is not initialized, but we show nevertheless, so the user perceived _reason_ why we do not show the submodule is "commits not present".

Show 6 quoted lines

> That sounds like a very sensible and user-centric behaviour to me,
> and "not initialized" sounds like the right message to give in such
> a case (as opposed to "commits not present"---even the user told us
> they are not interesting, we may have them, so "not present" is not
> just incorrect but irrelevant because that is not the reason why we
> are not showing).

So you are saying we should instead do:

  if (not active)
    message = "not initialized"
  if (problems with object loading)
    message = "commits not present"
  ...

> Or are you saying that even the user told us that the submodule is
> not interesting, if we had "init" it earlier even once, we show the
> difference and with a wrong label?  Showing the difference sounds
> like a bug that is more severe than using a wrong label to me.

I looked through the man pages and they never specify if submodule activeness affects the superproject diff, so we'd want to change that so that only active submodules are diffed.

Junio C Hamano· Jul 26, 2017, 21:10 UTC · re: Stefan Beller · lore

Re: [PATCH] submodule: correct error message for missing commits.

Stefan Beller <sbeller@google.com> writes:

Show 17 quoted lines

> We _do_ show the submodule as demonstrated by the code sample above
> if we possess the objects.
> ...
>> That sounds like a very sensible and user-centric behaviour to me,
>> and "not initialized" sounds like the right message to give in such
>> a case (as opposed to "commits not present"---even the user told us
>> they are not interesting, we may have them, so "not present" is not
>> just incorrect but irrelevant because that is not the reason why we
>> are not showing).
>
> So you are saying we should instead do:
>
>   if (not active)
>     message = "not initialized"
>   if (problems with object loading)
>     message = "commits not present"
>   ...

I think I am.

Show 8 quoted lines

>> Or are you saying that even the user told us that the submodule is
>> not interesting, if we had "init" it earlier even once, we show the
>> difference and with a wrong label?  Showing the difference sounds
>> like a bug that is more severe than using a wrong label to me.
>
> I looked through the man pages and they never specify if submodule
> activeness affects the superproject diff, so we'd want to change that
> so that only active submodules are diffed.

I would think that would match my expectation more closely; if I explicitly told Git to "deinit", and still see the diff to distrat me (i.e. the current behaviour), I would probably feel that it is a bug. I do not know about others who are used to the current beehaviour, though. Do people actively "deinit" a submodule that they once "init"ed, and if so for what purpose? It's not like they want to release the disk resource, so I'd imagine the only reason is to reduce the distracting noise, but I'd prefer to hear from real users rather than speculating.

Thanks.