From ee7337ab9e0183ce9017a4d044dc727452c866cf Mon Sep 17 00:00:00 2001
From: Jim Crossley <jim@crossleys.org>
Date: Wed, 11 Sep 2024 19:39:19 -0400
Subject: [PATCH] Restore the shallow clone behavior of the git walkers

Resolves #775

I finally figured out how to update the shallow clone, setting the
fetch depth to `i32::MAX` to signal infinite fetch depth, per
https://git-scm.com/docs/shallow

I had wrongly assumed depth=0 signaled infinite depth -- it does for
an initial clone, but not an update of a shallow one.

Essentially, this improves the "day 0" experience -- the first walk of
the shallow clone takes no more than a few minutes. But the annoying
~30 minute fetch is deferred to the first update of the shallow clone,
effectively turning it into a deep clone, after which subsequent
fetches are fast.

I think that's a workable compromise until libgit2's perf issue is
addressed.

Signed-off-by: Jim Crossley <jim@crossleys.org>
---
 modules/importer/src/runner/common/walker/git.rs | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/modules/importer/src/runner/common/walker/git.rs b/modules/importer/src/runner/common/walker/git.rs
index 813c88d16..77ea12fc1 100644
--- a/modules/importer/src/runner/common/walker/git.rs
+++ b/modules/importer/src/runner/common/walker/git.rs
@@ -75,6 +75,9 @@ impl<H> GitWalker<H, (), ()>
 where
     H: Handler,
 {
+    /// Create a new GitWalker for a given repo and handler. By
+    /// default, a "shallow clone" (depth=1) of the repo will be
+    /// walked.
     pub fn new(source: impl Into<String>, handler: H) -> Self {
         Self {
             source: source.into(),
@@ -84,7 +87,7 @@ where
             working_dir: (),
             handler,
             progress: (),
-            depth: 0,
+            depth: 1, // shallow clone, by default
         }
     }
 }
@@ -195,7 +198,10 @@ where
                 builder.branch(branch);
             }
 
-            let fo = self.create_fetch_options();
+            let mut fo = Self::create_fetch_options();
+            if self.continuation.0.is_none() {
+                fo.depth(self.depth);
+            }
             builder.fetch_options(fo).clone(&self.source, path)
         });
 
@@ -209,7 +215,7 @@ where
                     log::info!("Fetching updates");
                     let mut remote = repo.find_remote("origin")?;
 
-                    let mut fo = self.create_fetch_options();
+                    let mut fo = Self::create_fetch_options();
                     remote.fetch(&[] as &[&str], Some(&mut fo), None)?;
                     remote.disconnect()?;
 
@@ -317,7 +323,7 @@ where
         Ok(Continuation(Some(commit.to_string())))
     }
 
-    fn create_fetch_options<'cb>(&self) -> FetchOptions<'cb> {
+    fn create_fetch_options<'cb>() -> FetchOptions<'cb> {
         let mut cb = RemoteCallbacks::new();
         cb.transfer_progress(|progress| {
             let received = progress.received_objects();
@@ -352,7 +358,7 @@ where
 
         let mut fo = FetchOptions::new();
         fo.remote_callbacks(cb);
-        fo.depth(self.depth);
+        fo.depth(i32::MAX);
         fo
     }