This series improves the git submodule remote lookup logic implemented in submodule--helper. A few cleanups are done first: * Remove the branch->merge_name array and replace it by directly using branch->merge[i]->src immediately. This is simpler and easier to reason about. While cleaning this up, also fix the issues with branch_release() not tearing down everything properly. * remote_clear() failed to release the remote->push and remote->fetch refspec data. Fix this. * The starts_with_dot(_dot)_slash helper functions are moved to dir.h for re-use, as these are used both within submodule--helper.c and submodule-config.c * Several remote.c helper functions are refactored to take repository pointers, enabling use with a submodule repository pointer. Next, the submodule--helper.c logic replaces the repo_get_default_remote() function with a repo_default_remote() function in remote.c, which is based on the more robust configuration reading logic. This helper uses similar logic but also allows returning the only valid remote in the case where a repository has exactly one remote. This way we do not fall back to "origin" if a user has renamed the remote without adding another. This improved logic is a good first step, but won't handle cases where there are multiple remotes, with none of them being named "origin". For the final improvement, notice that the parent project already stores the URL for the submodule in its .git/config or .gitmodules file. This URL is what we use to set the remote in the first place when cloning. Add a repo_remote_from_url() helper which will iterate through the remotes and find the first remote with that URL. Use this in get_default_remote_submodule() to first try and find a remote by its URL. If unsuccessful, we still keep the original fallback logic, in the off chance that the user has changed the URL from within the submodule. This method is more robust as it is less likely that the user has manually changed the submodule URL within the submodule but not also within the .git/config. With this change, all commands which need the submodule remote will first look up by URL before trying to use the fallback logic, and should now be able to find a suitable remote regardless of now they are renamed. Signed-off-by: Jacob Keller <jacob.keller@xxxxxxxxx> --- Changes in v3: - Completely remove branch->merge_name, making the resulting logic much easier to understand. - Link to v2: https://lore.kernel.org/r/20250617-jk-submodule-helper-use-url-v2-0-04cbb003177d@xxxxxxxxx Changes in v2: - Remove repo_get_default_remote() entirely. The extra checks it does are really only necessary if you're doing manual configuration lookup. This avoids the confusion of similarly named functions and is less code. - Fix leaks in branch_release() and remote_clear(). - Add a forward declaration of struct repository. - Verified tests pass with leak sanitizer now. - Link to v1: https://lore.kernel.org/r/20250610-jk-submodule-helper-use-url-v1-0-6d14c1504e91@xxxxxxxxx --- Jacob Keller (7): remote: remove branch->merge_name and fix branch_release() remote: fix tear down of struct remote dir: move starts_with_dot(_dot)_slash to dir.h remote: remove the_repository from some functions submodule--helper: improve logic for fallback remote name submodule: move get_default_remote_submodule() submodule: look up remotes by URL first dir.h | 23 +++++++ remote.h | 8 ++- branch.c | 4 +- builtin/pull.c | 2 +- builtin/submodule--helper.c | 106 ++++++++++++-------------------- remote.c | 147 ++++++++++++++++++++++++++++---------------- submodule-config.c | 12 ---- t/t7406-submodule-update.sh | 61 ++++++++++++++++++ 8 files changed, 229 insertions(+), 134 deletions(-) --- base-commit: 16bd9f20a403117f2e0d9bcda6c6e621d3763e77 change-id: 20250610-jk-submodule-helper-use-url-e55d3c379faf Best regards, -- Jacob Keller <jacob.keller@xxxxxxxxx>
