hooks: add custom post-command hook config (#736)

dscho · dscho · commit ee7ed1424b57 · 2025-06-13T12:40:57.000+02:00
The microsoft/git fork includes pre- and post-command hooks, with the
initial intention of using these for VFS for Git. In that environment,
these are important hooks to avoid concurrent issues when the
virtualization is incomplete.

However, in the Office monorepo the post-command hook is used in a
different way. A custom hook is used to update the sparse-checkout, if
necessary. To avoid this hook from being incredibly slow on every Git
command, this hook checks for the existence of a "sentinel file" that is
written by a custom post-index-change hook and no-ops if that file does
not exist.

However, even this "no-op" is 200ms due to the use of two scripts (one
simple script in .git/hooks/ does some environment checking and then
calls a script from the working directory which actually contains the
logic).

Add a new config option, 'postCommand.strategy', that will allow for
multiple possible strategies in the future. For now, the one we are
adding is 'post-index-change' which states that we should write a
sentinel file instead of running the 'post-index-change' hook and then
skip the 'post-command' hook if the proper sentinel file doesn't exist.
(If it does exist, then delete it and run the hook.)

--- 

This fork contains changes specific to monorepo scenarios. If you are an
external contributor, then please detail your reason for submitting to
this fork:

* [ ] This is an early version of work already under review upstream.
* [ ] This change only applies to interactions with Azure DevOps and the
      GVFS Protocol.
* [ ] This change only applies to the virtualization hook and VFS for
Git.
* [x] This change only applies to custom bits in the microsoft/git fork.
diff --git a/Documentation/config.adoc b/Documentation/config.adoc
@@ -490,6 +490,8 @@ include::config/pack.adoc[]
 
 include::config/pager.adoc[]
 
+include::config/postcommand.adoc[]
+
 include::config/pretty.adoc[]
 
 include::config/promisor.adoc[]
diff --git a/Documentation/config/postcommand.adoc b/Documentation/config/postcommand.adoc
@@ -0,0 +1,13 @@
+postCommand.strategy::
+	The `post-command` hook is run on every Git process by default. This
+	config option allows running the hook only conditionally, according
+	to these values:
++
+----
+`always`;;
+	run the `post-command` hook on every process (default).
+
+`worktree-change`;;
+	run the `post-command` hook only if the current process wrote to
+	the index and updated the worktree.
+----
diff --git a/Documentation/config/protocol.adoc b/Documentation/config/protocol.adoc
@@ -6,7 +6,7 @@ protocol.allow::
 	default policy of `never`, and all other protocols (including file)
 	have a default policy of `user`.  Supported policies:
 +
---
+----
 
 * `always` - protocol is always able to be used.
 
@@ -18,15 +18,15 @@ protocol.allow::
   execute clone/fetch/push commands without user input, e.g. recursive
   submodule initialization.
 
---
+----
 
 protocol.<name>.allow::
 	Set a policy to be used by protocol `<name>` with clone/fetch/push
 	commands. See `protocol.allow` above for the available policies.
 +
 The protocol names currently used by git are:
 +
---
+----
   - `file`: any local file-based path (including `file://` URLs,
     or local paths)
 
@@ -42,7 +42,7 @@ The protocol names currently used by git are:
 
   - any external helpers are named by their protocol (e.g., use
     `hg` to allow the `git-remote-hg` helper)
---
+----
 
 protocol.version::
 	If set, clients will attempt to communicate with a server
@@ -51,7 +51,7 @@ protocol.version::
 	If unset, the default is `2`.
 	Supported versions:
 +
---
+----
 
 * `0` - the original wire protocol.
 
@@ -60,4 +60,4 @@ protocol.version::
 
 * `2` - Wire protocol version 2, see linkgit:gitprotocol-v2[5].
 
---
+----
diff --git a/Documentation/config/push.adoc b/Documentation/config/push.adoc
@@ -17,7 +17,7 @@ push.default::
 	(i.e. the fetch source is equal to the push destination),
 	`upstream` is probably what you want.  Possible values are:
 +
---
+----
 
 * `nothing` - do not push anything (error out) unless a refspec is
   given. This is primarily meant for people who want to
@@ -64,7 +64,7 @@ branches outside your control.
 This used to be the default, but not since Git 2.0 (`simple` is the
 new default).
 
---
+----
 
 push.followTags::
 	If set to true, enable `--follow-tags` option by default.  You
diff --git a/Documentation/config/sendemail.adoc b/Documentation/config/sendemail.adoc
@@ -58,7 +58,7 @@ the documentation of the email program of the same name. The
 differences and limitations from the standard formats are
 described below:
 +
---
+----
 sendmail;;
 *	Quoted aliases and quoted addresses are not supported: lines that
 	contain a `"` symbol are ignored.
@@ -68,7 +68,7 @@ sendmail;;
 *	Warnings are printed on the standard error output for any
 	explicitly unsupported constructs, and any other lines that are not
 	recognized by the parser.
---
+----
 sendemail.annotate::
 sendemail.bcc::
 sendemail.cc::
diff --git a/Documentation/config/sideband.adoc b/Documentation/config/sideband.adoc
@@ -4,7 +4,7 @@ sideband.allowControlCharacters::
 	unwanted ANSI escape sequences from being sent to the terminal. Use
 	this config setting to override this behavior:
 +
---
+----
 	color::
 		Allow ANSI color sequences, line feeds and horizontal tabs,
 		but mask all other control characters. This is the default.
@@ -13,4 +13,4 @@ sideband.allowControlCharacters::
 		horizontal tabs.
 	true::
 		Allow all control characters to be sent to the terminal.
---
+----
diff --git a/Documentation/config/ssh.adoc b/Documentation/config/ssh.adoc
@@ -19,7 +19,7 @@ overridden via the environment variable `GIT_SSH_VARIANT`.
 The current command-line parameters used for each variant are as
 follows:
 +
---
+----
 
 * `ssh` - [-p port] [-4] [-6] [-o option] [username@]host command
 
@@ -29,7 +29,7 @@ follows:
 
 * `tortoiseplink` - [-P port] [-4] [-6] -batch [username@]host command
 
---
+----
 +
 Except for the `simple` variant, command-line parameters are likely to
 change as git gains new features.
diff --git a/Documentation/config/status.adoc b/Documentation/config/status.adoc
@@ -50,11 +50,11 @@ status.showUntrackedFiles::
 	systems. So, this variable controls how the commands display
 	the untracked files. Possible values are:
 +
---
+----
 * `no` - Show no untracked files.
 * `normal` - Show untracked files and directories.
 * `all` - Show also individual files in untracked directories.
---
+----
 +
 If this variable is not specified, it defaults to 'normal'.
 All usual spellings for Boolean value `true` are taken as `normal`
@@ -90,12 +90,12 @@ status.deserializeWait::
 	fall-back and compute status normally.  This will be overridden by
 	`--deserialize-wait=<value>` on the command line.
 +
---
+----
 * `fail` - cause git to exit with an error when the status cache file
 is stale; this is intended for testing and debugging.
 * `block` - cause git to spin and periodically retry the cache file
 every 100 ms; this is intended to help coordinate with another git
 instance concurrently computing the cache file.
 * `no` - to immediately fall-back if cache file is stale.  This is the default.
 * `<timeout>` - time (in tenths of a second) to spin and retry.
---
+----
diff --git a/Documentation/config/survey.adoc b/Documentation/config/survey.adoc
@@ -3,7 +3,7 @@ survey.*::
 	command. The intention is that this command could be run in the
 	background with these options.
 +
---
+----
 	survey.namerev::
 		Boolean to show/hide `git name-rev` information for each
 		reported commit and the containing commit of each
@@ -44,4 +44,4 @@ survey.*::
 		long file or subdirectory entry names.  Provides a
 		default value for `--tree-sizes=<n>` in
 		linkgit:git-survey[1].
---
+----
diff --git a/Documentation/config/trace2.adoc b/Documentation/config/trace2.adoc
@@ -16,7 +16,7 @@ trace2.eventTarget::
 	This variable controls the event target destination.
 	It may be overridden by the `GIT_TRACE2_EVENT` environment variable.
 	The following table shows possible values.
-+
+
 include::../trace2-target-values.adoc[]
 
 trace2.normalBrief::
diff --git a/Documentation/trace2-target-values.adoc b/Documentation/trace2-target-values.adoc
@@ -1,12 +1,12 @@
---
+----
 * `0` or `false` - Disables the target.
 * `1` or `true` - Writes to `STDERR`.
 * `[2-9]` - Writes to the already opened file descriptor.
 * `<absolute-pathname>` - Writes to the file in append mode. If the target
-already exists and is a directory, the traces will be written to files (one
-per process) underneath the given directory.
+  already exists and is a directory, the traces will be written to files (one
+  per process) underneath the given directory.
 * `af_unix:[<socket-type>:]<absolute-pathname>` - Write to a
-Unix DomainSocket (on platforms that support them).  Socket
-type can be either `stream` or `dgram`; if omitted Git will
-try both.
---
+  Unix DomainSocket (on platforms that support them).  Socket
+  type can be either `stream` or `dgram`; if omitted Git will
+  try both.
+----
diff --git a/git.c b/git.c
@@ -495,6 +495,7 @@ static int run_post_command_hook(struct repository *r)
 
 	run_post_hook = 0;
 	strvec_clear(&sargv);
+	strvec_clear(&opt.args);
 	setenv("COMMAND_HOOK_LOCK", "false", 1);
 	return ret;
 }
diff --git a/hook.c b/hook.c
@@ -1,6 +1,7 @@
 #define USE_THE_REPOSITORY_VARIABLE
 
 #include "git-compat-util.h"
+#include "trace2/tr2_sid.h"
 #include "abspath.h"
 #include "environment.h"
 #include "advice.h"
@@ -176,6 +177,93 @@ static void run_hooks_opt_clear(struct run_hooks_opt *options)
 	strvec_clear(&options->args);
 }
 
+static char *get_post_index_change_sentinel_name(struct repository *r)
+{
+	struct strbuf path = STRBUF_INIT;
+	const char *sid = tr2_sid_get();
+	char *slash = strchr(sid, '/');
+
+	/*
+	 * Name is based on top-level SID, so children can indicate that
+	 * the top-level process should run the post-command hook.
+	 */
+	if (slash)
+		*slash = 0;
+
+	/*
+	 * Do not write to hooks directory, as it could be redirected
+	 * somewhere like the source tree.
+	 */
+	repo_git_path_replace(r, &path, "info/index-change-%s.snt", sid);
+
+	return strbuf_detach(&path, NULL);
+}
+
+static int write_post_index_change_sentinel(struct repository *r)
+{
+	char *path = get_post_index_change_sentinel_name(r);
+	FILE *fp = xfopen(path, "w");
+
+	if (fp) {
+		fprintf(fp, "run post-command hook");
+		fclose(fp);
+	}
+
+	free(path);
+	return fp ? 0 : -1;
+}
+
+/**
+ * Try to delete the sentinel file for this repository. If that succeeds, then
+ * return 1.
+ */
+static int post_index_change_sentinel_exists(struct repository *r)
+{
+	char *path = get_post_index_change_sentinel_name(r);
+	int res = 1;
+
+	if (unlink(path)) {
+		if (is_missing_file_error(errno))
+			res = 0;
+		else
+			warning_errno("failed to remove index-change sentinel file '%s'", path);
+	}
+
+	free(path);
+	return res;
+}
+
+/**
+ * See if we can replace the requested hook with an internal behavior.
+ * Returns 0 if the real hook should run. Returns nonzero if we instead
+ * executed custom internal behavior and the real hook should not run.
+ */
+static int handle_hook_replacement(struct repository *r,
+				   const char *hook_name,
+				   struct strvec *args)
+{
+	const char *strval;
+	if (repo_config_get_string_tmp(r, "postcommand.strategy", &strval) ||
+	    strcasecmp(strval, "worktree-change"))
+		return 0;
+
+	if (!strcmp(hook_name, "post-index-change")) {
+		/* Create a sentinel file only if the worktree changed. */
+		if (!strcmp(args->v[0], "1"))
+			write_post_index_change_sentinel(r);
+
+		/* We don't skip post-index-change hooks that exist. */
+		return 0;
+	}
+	if (!strcmp(hook_name, "post-command") &&
+	    !post_index_change_sentinel_exists(r)) {
+		/* We skip the post-command hook in this case. */
+		return 1;
+	}
+
+	return 0;
+}
+
 int run_hooks_opt(struct repository *r, const char *hook_name,
 		  struct run_hooks_opt *options)
 {
@@ -185,7 +273,7 @@ int run_hooks_opt(struct repository *r, const char *hook_name,
 		.hook_name = hook_name,
 		.options = options,
 	};
-	const char *hook_path = find_hook(r, hook_name);
+	const char *hook_path;
 	int ret = 0;
 	const struct run_process_parallel_opts opts = {
 		.tr2_category = "hook",
@@ -201,6 +289,13 @@ int run_hooks_opt(struct repository *r, const char *hook_name,
 		.data = &cb_data,
 	};
 
+	/* Interject hook behavior depending on strategy. */
+	if (r && r->gitdir &&
+	    handle_hook_replacement(r, hook_name, &options->args))
+		return 0;
+
+	hook_path = find_hook(r, hook_name);
+
 	/*
 	 * Backwards compatibility hack in VFS for Git: when originally
 	 * introduced (and used!), it was called `post-indexchanged`, but this
@@ -252,12 +347,15 @@ int run_hooks_l(struct repository *r, const char *hook_name, ...)
 {
 	struct run_hooks_opt opt = RUN_HOOKS_OPT_INIT;
 	va_list ap;
+	int result;
 	const char *arg;
 
 	va_start(ap, hook_name);
 	while ((arg = va_arg(ap, const char *)))
 		strvec_push(&opt.args, arg);
 	va_end(ap);
 
-	return run_hooks_opt(r, hook_name, &opt);
+	result = run_hooks_opt(r, hook_name, &opt);
+	strvec_clear(&opt.args);
+	return result;
 }
diff --git a/t/t0401-post-command-hook.sh b/t/t0401-post-command-hook.sh

Original file line number	Diff line number	Diff line change
@@ -19,7 +19,7 @@ overridden via the environment variable `GIT_SSH_VARIANT`.
`19`	`19`	`The current command-line parameters used for each variant are as`
`20`	`20`	`follows:`
`21`	`21`	`+`
`22`		`---`
	`22`	`+----`
`23`	`23`
`24`	`24`	* `ssh` - [-p port] [-4] [-6] [-o option] [username@]host command
`25`	`25`
`@@ -29,7 +29,7 @@ follows:`
`29`	`29`
`30`	`30`	* `tortoiseplink` - [-P port] [-4] [-6] -batch [username@]host command
`31`	`31`
`32`		`---`
	`32`	`+----`
`33`	`33`	`+`
`34`	`34`	Except for the `simple` variant, command-line parameters are likely to
`35`	`35`	`change as git gains new features.`