working_copy: don't snapshot new files larger than 1MiB by default

CHANGELOG.md

@@ -29,6 +29,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 * The default editor on Windows is now `Notepad` instead of `pico`.
 
+* `jj` will fail attempts to snapshot new files larger than 1MiB by default. This behavior
+  can be customized with the `snapshot.max-new-file-size` config option.
+
+
 ### New features
 
 * Default template for `jj log` now does not show irrelevant information

cli/src/cli_util.rs

@@ -1289,6 +1289,7 @@ See https://github.com/martinvonz/jj/blob/main/docs/working-copy.md#stale-workin
             base_ignores,
             fsmonitor_kind: self.settings.fsmonitor_kind()?,
             progress: progress.as_ref().map(|x| x as _),
+            max_new_file_size: self.settings.max_new_file_size()?,
         })?;
         drop(progress);
         if new_tree_id != *wc_commit.tree_id() {

cli/src/commands/mod.rs

@@ -1338,6 +1338,7 @@ fn cmd_untrack(
         base_ignores,
         fsmonitor_kind: command.settings().fsmonitor_kind()?,
         progress: None,
+        max_new_file_size: command.settings().max_new_file_size()?,
     })?;
     if wc_tree_id != new_tree_id {
         let wc_tree = store.get_tree(&RepoPath::root(), &wc_tree_id)?;

cli/src/config-schema.json

@@ -329,6 +329,17 @@
                     "type": "string"
                 }
             }
+        },
+        "snapshot": {
+            "type": "object",
+            "description": "Parameters governing automatic capture of files into the working copy commit",
+            "properties": {
+                "max-new-file-size": {
+                    "type": "integer",
+                    "description": "New files with a size in bytes above this threshold are not snapshotted, unless the threshold is 0",
+                    "default": "1048576"
+                }
+            }
         }
     }
 }

cli/src/config/misc.toml

@@ -8,3 +8,6 @@
 paginate = "auto"
 pager = { command = ["less", "-FRX"], env = { LESSCHARSET = "utf-8" } }
 log-word-wrap = false
+
+[snapshot]
+max-new-file-size = 1048576

cli/src/merge_tools/external.rs

@@ -414,6 +414,7 @@ pub fn edit_diff_external(
         base_ignores,
         fsmonitor_kind: settings.fsmonitor_kind()?,
         progress: None,
+        max_new_file_size: settings.max_new_file_size()?,
     })?;
     Ok(right_tree_state.current_tree_id().clone())
 }

lib/src/settings.rs

@@ -196,6 +196,16 @@ impl UserSettings {
             .get_string("ui.graph.style")
             .unwrap_or_else(|_| "curved".to_string())
     }
+
+    pub fn max_new_file_size(&self) -> Result<u64, config::ConfigError> {
+        let cfg = self.config.get::<u64>("snapshot.max-new-file-size");
+        match cfg {
+            Ok(0) => Ok(u64::MAX),
+            x @ Ok(_) => x,
+            Err(config::ConfigError::NotFound(_)) => Ok(1024 * 1024),
+            e @ Err(_) => e,
+        }
+    }
 }
 
 /// This Rng uses interior mutability to allow generating random values using an

lib/src/working_copy.rs

@@ -310,6 +310,12 @@ pub enum SnapshotError {
     InternalBackendError(#[from] BackendError),
     #[error(transparent)]
     TreeStateError(#[from] TreeStateError),
+    #[error("New file {path} of size {size} exceeds snapshot.max-new-file-size ({max_size})")]
+    NewFileTooLarge {
+        path: PathBuf,
+        size: u64,
+        max_size: u64,
+    },
 }
 
 #[derive(Debug, Error)]
@@ -358,6 +364,7 @@ pub struct SnapshotOptions<'a> {
     pub base_ignores: Arc<GitIgnoreFile>,
     pub fsmonitor_kind: Option<FsmonitorKind>,
     pub progress: Option<&'a SnapshotProgress<'a>>,
+    pub max_new_file_size: u64,
 }
 
 impl SnapshotOptions<'_> {
@@ -366,6 +373,7 @@ impl SnapshotOptions<'_> {
             base_ignores: GitIgnoreFile::empty(),
             fsmonitor_kind: None,
             progress: None,
+            max_new_file_size: u64::MAX,
         }
     }
 }
@@ -630,6 +638,7 @@ impl TreeState {
             base_ignores,
             fsmonitor_kind,
             progress,
+            max_new_file_size,
         } = options;
 
         let sparse_matcher = self.sparse_matcher();
@@ -665,6 +674,7 @@ impl TreeState {
                 present_files_tx,
                 directory_to_visit,
                 progress,
+                max_new_file_size,
             )
         })?;
 
@@ -736,6 +746,7 @@ impl TreeState {
         present_files_tx: Sender<RepoPath>,
         directory_to_visit: DirectoryToVisit,
         progress: Option<&SnapshotProgress>,
+        max_new_file_size: u64,
     ) -> Result<(), SnapshotError> {
         let DirectoryToVisit {
             dir,
@@ -841,6 +852,7 @@ impl TreeState {
                             present_files_tx.clone(),
                             directory_to_visit,
                             progress,
+                            max_new_file_size,
                         )?;
                     }
                 } else if matcher.matches(&path) {
@@ -859,6 +871,14 @@ impl TreeState {
                             message: format!("Failed to stat file {}", entry.path().display()),
                             err,
                         })?;
+                        if maybe_current_file_state.is_none() && metadata.len() > max_new_file_size
+                        {
+                            return Err(SnapshotError::NewFileTooLarge {
+                                path: entry.path().clone(),
+                                size: metadata.len(),
+                                max_size: max_new_file_size,
+                            });
+                        }
                         if let Some(new_file_state) = file_state(&metadata) {
                             present_files_tx.send(path.clone()).ok();
                             let update = self.get_updated_tree_value(