Re: [PATCH] submodule: mark submodules with update=none as inactive

Hi brian,

Le 2021-06-25 à 19:02, brian m. carlson a écrit :

On 2021-06-22 at 03:45:45, Philippe Blain wrote:

quoted

quoted

That will make us properly ignore the submodule
when performing recursive operations.

Note that we only do this when the --require-init option is passed,
which is only passed during clone.  That's because the update-clone
submodule helper is also invoked during a user-invoked git submodule
update, where we explicitly indicate that we override the configuration
setting, so we don't want to set such a configuration option then.

I'm not sure what you mean here by 'where we explicitely indicate that we
override the configuration setting'. For me, as I wrote above,
'git clone --recurse-submodules' and 'git clone' followed by
'git submodule update --init' should lead to mostly [*] the same end result.

If you mean 'git submodule update --checkout', that indeed seems to sometimes override the 'update=none'
configuration (a fact which is absent from the documentation), 

Note that I'm taking back that statemnt about the doc; the description of 'git submodule update'
states:

"The "updating" can be done in several ways depending on command line options
and the value of submodule.<name>.update configuration variable. The command line
option takes precedence over the configuration variable."

I was somehow under the impression that 'none' was a special case, but it's not.

then it's true that we

quoted

would not want to write 'active=false' at that invocation. As an aside, in my limited testing
I could not always get 'git submodule update --checkout' to clone and checkout 'update=none' submdules;
it would fail with "fatal: could not get a repository handle for submodule 'sub1'" because
'git checkout/reset --recurse-submodules' leaves a bad .git/modules/sub1/config file
with the core.worktree setting when the command fails (this should not happen)...

Yes, that's what I meant.

quoted

In any case, that leads me to think that maybe the right place to write the 'active' setting
would be during 'git submodule init', thus builtin/submodule--helper.c::init_submodule ?
This way it would lead to the same behaviour if the clone was recursive or not,
and it would not interfere with 'git submodule update --checkout'.

Let me take a look at some other approaches and see if I can come up
with something a little bit better.

I tested the following and it fixes the bug (for both recursive clone and
normal clone followed by 'git submodule update --init') and
t7406-submodule-update.sh still passes:

diff --git a/builtin/submodule--helper.c b/builtin/submodule--helper.c
index d55f6262e9..a4cd86c72f 100644
--- a/builtin/submodule--helper.c
+++ b/builtin/submodule--helper.c

@@ -686,6 +686,13 @@ static void init_submodule(const char *path, const char *prefix,
  
  		if (git_config_set_gently(sb.buf, upd))
  			die(_("Failed to register update mode for submodule path '%s'"), displaypath);
+
+		/* Mark update=none submodules as inactive */
+		if (sub->update_strategy.type == SM_UPDATE_NONE) {
+			strbuf_reset(&sb);
+			strbuf_addf(&sb, "submodule.%s.active", sub->name);
+			git_config_set_gently(sb.buf, "false");
+		}
  	}
  	strbuf_release(&sb);
  	free(displaypath);

My apologies for the delay in response; I'm in the process of moving at
the moment and my attention has been directed elsewhere than the list.

No problem of course!

Philippe.


[1] https://git-scm.com/docs/git-submodule#Documentation/git-submodule.txt-update--init--remote-N--no-fetch--no-recommend-shallow-f--force--checkout--rebase--merge--referenceltrepositorygt--depthltdepthgt--recursive--jobsltngt--no-single-branch--ltpathgt82308203