Bug 536534

Summary:	git-r3.eclass should support fetching repository with submodule with no HEAD
Product:	Gentoo Linux	Reporter:	Denis <public>
Component:	Eclasses	Assignee:	Michał Górny <mgorny>
Status:	RESOLVED NEEDINFO
Severity:	normal
Priority:	Normal
Version:	unspecified
Hardware:	All
OS:	Linux
See Also:	https://bugs.gentoo.org/show_bug.cgi?id=524146
Whiteboard:
Package list:		Runtime testing required:	---

Description Denis 2015-01-13 21:23:17 UTC

Bare git repositories may have no HEAD, it happens sometimes.

It's ok when we have such repository in our EGIT_REPO_URI since it's enough to set some valid EGIT_BRANCH to avoid a fetching error "fatal: Couldn't find remote ref HEAD". It works even if we want to work with the particular commit and have set EGIT_COMMIT -- they work together perferctly.

The problem raises when the main repository has a submodule and this submodule's repository doesn't have any HEAD. This is what git-r3 does not expect and it doesn't allow to eigher turn off submodule loading (if it's not needed) or to set some branch to be looked at instead of the HEAD.

So, I propose to change the logic of fetching in case fetched repository has no HEAD but still has some branch.

Comment 1 Michał Górny archtester

2016-05-25 17:34:05 UTC

Could you point me to an ebuild I could use to test this?

Comment 2 Michał Górny archtester

2016-05-25 17:41:11 UTC

Possibly the fix from bug #524146 could solve this as well.

Comment 3 Michał Górny archtester

2016-05-26 10:25:55 UTC

Expanding on this, it looks weird to me. AFAIR HEAD on a bare repository specifies the default branch to fetch, so what is fetched when you want to clone it? Does git have some implicit fallback to 'master'? Should we maybe do a full mirror-style clone in this scenario?

Comment 4 Michał Górny archtester

2016-10-22 20:57:00 UTC

I'm closing due to lack of reply. Please reopen when you can provide the details needed for testing this.