Here's some code that performs the lazy fetch in a subprocess. It was more involved than I thought it would be, but here it is. My main concern is the user-facing interface for the null fetch negotiator. It not only does not negotiate at all but affects the everything_local() check, so I wonder if it should be exposed differently. More information is in the commit message for the 2nd patch. I know that the following still needs to be done, but I thought I'd send out the patches for early feedback first since the main design and code is already done: - Commit messages - User-facing documentation - A way to prevent a promisor-remote fetch from invoking another promisor-remote fetch (use a file as a lock?) - Remove no_dependents code (fetch-pack, transport) Jonathan Tan (7): fetch-pack: allow NULL negotiator->add_tip fetch-pack: allow NULL negotiator->known_common negotiator/null: add null fetch negotiator fetch: --stdin fetch: submodule config fetch: only populate existing_refs if needed promisor-remote: use subprocess to fetch Documentation/config/fetch.txt | 5 +++- Makefile | 1 + builtin/fetch.c | 42 +++++++++++++++++++++++------ fetch-negotiator.c | 5 ++++ fetch-pack.c | 23 +++++++++++----- negotiator/null.c | 34 +++++++++++++++++++++++ negotiator/null.h | 8 ++++++ promisor-remote.c | 46 +++++++++++++++----------------- repo-settings.c | 2 ++ repository.h | 1 + submodule-config.c | 5 ++-- t/t0410-partial-clone.sh | 2 +- t/t4067-diff-partial-clone.sh | 8 +++--- t/t5554-null-fetch-negotiator.sh | 22 +++++++++++++++ t/t5601-clone.sh | 2 +- 15 files changed, 157 insertions(+), 49 deletions(-) create mode 100644 negotiator/null.c create mode 100644 negotiator/null.h create mode 100755 t/t5554-null-fetch-negotiator.sh -- 2.28.0.163.g6104cc2f0b6-goog
