From 78db6ea9dc1a872f9d07a36fe7aec700a5c963b9 Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:18:29 -0500
Subject: grep: make locking flag global

The low-level grep code traditionally didn't care about
threading, as it doesn't do any threading itself and didn't
call out to other non-thread-safe code.  That changed with
0579f91 (grep: enable threading with -p and -W using lazy
attribute lookup, 2011-12-12), which pushed the lookup of
funcname attributes (which is not thread-safe) into the
low-level grep code.

As a result, the low-level code learned about a new global
"grep_attr_mutex" to serialize access to the attribute code.
A multi-threaded caller (e.g., builtin/grep.c) is expected
to initialize the mutex and set "use_threads" in the
grep_opt structure. The low-level code only uses the lock if
use_threads is set.

However, putting the use_threads flag into the grep_opt
struct is not the most logical place. Whether threading is
in use is not something that matters for each call to
grep_buffer, but is instead global to the whole program
(i.e., if any thread is doing multi-threaded grep, every
other thread, even if it thinks it is doing its own
single-threaded grep, would need to use the locking).  In
practice, this distinction isn't a problem for us, because
the only user of multi-threaded grep is "git-grep", which
does nothing except call grep.

This patch turns the opt->use_threads flag into a global
flag. More important than the nit-picking semantic argument
above is that this means that the locking functions don't
need to actually have access to a grep_opt to know whether
to lock. Which in turn can make adding new locks simpler, as
we don't need to pass around a grep_opt.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/builtin/grep.c b/builtin/grep.c
index 9ce064a..5b88187 100644
--- a/builtin/grep.c
+++ b/builtin/grep.c
@@ -259,6 +259,7 @@ static void start_threads(struct grep_opt *opt)
 	pthread_cond_init(&cond_add, NULL);
 	pthread_cond_init(&cond_write, NULL);
 	pthread_cond_init(&cond_result, NULL);
+	grep_use_locks = 1;
 
 	for (i = 0; i < ARRAY_SIZE(todo); i++) {
 		strbuf_init(&todo[i].out, 0);
@@ -307,6 +308,7 @@ static int wait_all(void)
 	pthread_cond_destroy(&cond_add);
 	pthread_cond_destroy(&cond_write);
 	pthread_cond_destroy(&cond_result);
+	grep_use_locks = 0;
 
 	return hit;
 }
@@ -1030,8 +1032,6 @@ int cmd_grep(int argc, const char **argv, const char *prefix)
 	use_threads = 0;
 #endif
 
-	opt.use_threads = use_threads;
-
 #ifndef NO_PTHREADS
 	if (use_threads) {
 		if (opt.pre_context || opt.post_context || opt.file_break ||
diff --git a/grep.c b/grep.c
index 486230b..7a67c2f 100644
--- a/grep.c
+++ b/grep.c
@@ -807,26 +807,28 @@ static void show_line(struct grep_opt *opt, char *bol, char *eol,
 }
 
 #ifndef NO_PTHREADS
+int grep_use_locks;
+
 /*
  * This lock protects access to the gitattributes machinery, which is
  * not thread-safe.
  */
 pthread_mutex_t grep_attr_mutex;
 
-static inline void grep_attr_lock(struct grep_opt *opt)
+static inline void grep_attr_lock(void)
 {
-	if (opt->use_threads)
+	if (grep_use_locks)
 		pthread_mutex_lock(&grep_attr_mutex);
 }
 
-static inline void grep_attr_unlock(struct grep_opt *opt)
+static inline void grep_attr_unlock(void)
 {
-	if (opt->use_threads)
+	if (grep_use_locks)
 		pthread_mutex_unlock(&grep_attr_mutex);
 }
 #else
-#define grep_attr_lock(opt)
-#define grep_attr_unlock(opt)
+#define grep_attr_lock()
+#define grep_attr_unlock()
 #endif
 
 static int match_funcname(struct grep_opt *opt, const char *name, char *bol, char *eol)
@@ -834,9 +836,9 @@ static int match_funcname(struct grep_opt *opt, const char *name, char *bol, cha
 	xdemitconf_t *xecfg = opt->priv;
 	if (xecfg && !xecfg->find_func) {
 		struct userdiff_driver *drv;
-		grep_attr_lock(opt);
+		grep_attr_lock();
 		drv = userdiff_find_by_path(name);
-		grep_attr_unlock(opt);
+		grep_attr_unlock();
 		if (drv && drv->funcname.pattern) {
 			const struct userdiff_funcname *pe = &drv->funcname;
 			xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
diff --git a/grep.h b/grep.h
index fb205f3..3653bb3 100644
--- a/grep.h
+++ b/grep.h
@@ -116,7 +116,6 @@ struct grep_opt {
 	int show_hunk_mark;
 	int file_break;
 	int heading;
-	int use_threads;
 	void *priv;
 
 	void (*output)(struct grep_opt *opt, const void *data, size_t size);
@@ -138,6 +137,7 @@ extern int grep_threads_ok(const struct grep_opt *opt);
  * Mutex used around access to the attributes machinery if
  * opt->use_threads.  Must be initialized/destroyed by callers!
  */
+extern int grep_use_locks;
 extern pthread_mutex_t grep_attr_mutex;
 #endif
 
-- 
cgit v0.10.2-6-g49f6


From b3aeb285d0ac1dcb4d578a61a68e08646f96501f Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:18:41 -0500
Subject: grep: move sha1-reading mutex into low-level code

The multi-threaded git-grep code needs to serialize access
to the thread-unsafe read_sha1_file call. It does this with
a mutex that is local to builtin/grep.c.

Let's instead push this down into grep.c, where it can be
used by both builtin/grep.c and grep.c. This will let us
safely teach the low-level grep.c code tricks that involve
reading from the object db.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/builtin/grep.c b/builtin/grep.c
index 5b88187..aeb3616 100644
--- a/builtin/grep.c
+++ b/builtin/grep.c
@@ -85,21 +85,6 @@ static inline void grep_unlock(void)
 		pthread_mutex_unlock(&grep_mutex);
 }
 
-/* Used to serialize calls to read_sha1_file. */
-static pthread_mutex_t read_sha1_mutex;
-
-static inline void read_sha1_lock(void)
-{
-	if (use_threads)
-		pthread_mutex_lock(&read_sha1_mutex);
-}
-
-static inline void read_sha1_unlock(void)
-{
-	if (use_threads)
-		pthread_mutex_unlock(&read_sha1_mutex);
-}
-
 /* Signalled when a new work_item is added to todo. */
 static pthread_cond_t cond_add;
 
@@ -254,7 +239,7 @@ static void start_threads(struct grep_opt *opt)
 	int i;
 
 	pthread_mutex_init(&grep_mutex, NULL);
-	pthread_mutex_init(&read_sha1_mutex, NULL);
+	pthread_mutex_init(&grep_read_mutex, NULL);
 	pthread_mutex_init(&grep_attr_mutex, NULL);
 	pthread_cond_init(&cond_add, NULL);
 	pthread_cond_init(&cond_write, NULL);
@@ -303,7 +288,7 @@ static int wait_all(void)
 	}
 
 	pthread_mutex_destroy(&grep_mutex);
-	pthread_mutex_destroy(&read_sha1_mutex);
+	pthread_mutex_destroy(&grep_read_mutex);
 	pthread_mutex_destroy(&grep_attr_mutex);
 	pthread_cond_destroy(&cond_add);
 	pthread_cond_destroy(&cond_write);
@@ -313,8 +298,6 @@ static int wait_all(void)
 	return hit;
 }
 #else /* !NO_PTHREADS */
-#define read_sha1_lock()
-#define read_sha1_unlock()
 
 static int wait_all(void)
 {
@@ -376,9 +359,9 @@ static void *lock_and_read_sha1_file(const unsigned char *sha1, enum object_type
 {
 	void *data;
 
-	read_sha1_lock();
+	grep_read_lock();
 	data = read_sha1_file(sha1, type, size);
-	read_sha1_unlock();
+	grep_read_unlock();
 	return data;
 }
 
@@ -617,10 +600,10 @@ static int grep_object(struct grep_opt *opt, const struct pathspec *pathspec,
 		struct strbuf base;
 		int hit, len;
 
-		read_sha1_lock();
+		grep_read_lock();
 		data = read_object_with_reference(obj->sha1, tree_type,
 						  &size, NULL);
-		read_sha1_unlock();
+		grep_read_unlock();
 
 		if (!data)
 			die(_("unable to read tree (%s)"), sha1_to_hex(obj->sha1));
diff --git a/grep.c b/grep.c
index 7a67c2f..db58a29 100644
--- a/grep.c
+++ b/grep.c
@@ -826,6 +826,12 @@ static inline void grep_attr_unlock(void)
 	if (grep_use_locks)
 		pthread_mutex_unlock(&grep_attr_mutex);
 }
+
+/*
+ * Same as git_attr_mutex, but protecting the thread-unsafe object db access.
+ */
+pthread_mutex_t grep_read_mutex;
+
 #else
 #define grep_attr_lock()
 #define grep_attr_unlock()
diff --git a/grep.h b/grep.h
index 3653bb3..4f1b025 100644
--- a/grep.h
+++ b/grep.h
@@ -139,6 +139,23 @@ extern int grep_threads_ok(const struct grep_opt *opt);
  */
 extern int grep_use_locks;
 extern pthread_mutex_t grep_attr_mutex;
+extern pthread_mutex_t grep_read_mutex;
+
+static inline void grep_read_lock(void)
+{
+	if (grep_use_locks)
+		pthread_mutex_lock(&grep_read_mutex);
+}
+
+static inline void grep_read_unlock(void)
+{
+	if (grep_use_locks)
+		pthread_mutex_unlock(&grep_read_mutex);
+}
+
+#else
+#define grep_read_lock()
+#define grep_read_unlock()
 #endif
 
 #endif
-- 
cgit v0.10.2-6-g49f6


From e1327023ea22c3bf57e7d28596da356043f073fc Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:19:28 -0500
Subject: grep: refactor the concept of "grep source" into an object

The main interface to the low-level grep code is
grep_buffer, which takes a pointer to a buffer and a size.
This is convenient and flexible (we use it to grep commit
bodies, files on disk, and blobs by sha1), but it makes it
hard to pass extra information about what we are grepping
(either for correctness, like overriding binary
auto-detection, or for optimizations, like lazily loading
blob contents).

Instead, let's encapsulate the idea of a "grep source",
including the buffer, its size, and where the data is coming
from. This is similar to the diff_filespec structure used by
the diff code (unsurprising, since future patches will
implement some of the same optimizations found there).

The diffstat is slightly scarier than the actual patch
content. Most of the modified lines are simply replacing
access to raw variables with their counterparts that are now
in a "struct grep_source". Most of the added lines were
taken from builtin/grep.c, which partially abstracted the
idea of grep sources (for file vs sha1 sources).

Instead of dropping the now-redundant code, this patch
leaves builtin/grep.c using the traditional grep_buffer
interface (which now wraps the grep_source interface). That
makes it easy to test that there is no change of behavior
(yet).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/grep.c b/grep.c
index db58a29..8204ca2 100644
--- a/grep.c
+++ b/grep.c
@@ -837,13 +837,13 @@ pthread_mutex_t grep_read_mutex;
 #define grep_attr_unlock()
 #endif
 
-static int match_funcname(struct grep_opt *opt, const char *name, char *bol, char *eol)
+static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bol, char *eol)
 {
 	xdemitconf_t *xecfg = opt->priv;
 	if (xecfg && !xecfg->find_func) {
 		struct userdiff_driver *drv;
 		grep_attr_lock();
-		drv = userdiff_find_by_path(name);
+		drv = userdiff_find_by_path(gs->name);
 		grep_attr_unlock();
 		if (drv && drv->funcname.pattern) {
 			const struct userdiff_funcname *pe = &drv->funcname;
@@ -866,33 +866,33 @@ static int match_funcname(struct grep_opt *opt, const char *name, char *bol, cha
 	return 0;
 }
 
-static void show_funcname_line(struct grep_opt *opt, const char *name,
-			       char *buf, char *bol, unsigned lno)
+static void show_funcname_line(struct grep_opt *opt, struct grep_source *gs,
+			       char *bol, unsigned lno)
 {
-	while (bol > buf) {
+	while (bol > gs->buf) {
 		char *eol = --bol;
 
-		while (bol > buf && bol[-1] != '\n')
+		while (bol > gs->buf && bol[-1] != '\n')
 			bol--;
 		lno--;
 
 		if (lno <= opt->last_shown)
 			break;
 
-		if (match_funcname(opt, name, bol, eol)) {
-			show_line(opt, bol, eol, name, lno, '=');
+		if (match_funcname(opt, gs, bol, eol)) {
+			show_line(opt, bol, eol, gs->name, lno, '=');
 			break;
 		}
 	}
 }
 
-static void show_pre_context(struct grep_opt *opt, const char *name, char *buf,
+static void show_pre_context(struct grep_opt *opt, struct grep_source *gs,
 			     char *bol, char *end, unsigned lno)
 {
 	unsigned cur = lno, from = 1, funcname_lno = 0;
 	int funcname_needed = !!opt->funcname;
 
-	if (opt->funcbody && !match_funcname(opt, name, bol, end))
+	if (opt->funcbody && !match_funcname(opt, gs, bol, end))
 		funcname_needed = 2;
 
 	if (opt->pre_context < lno)
@@ -901,14 +901,14 @@ static void show_pre_context(struct grep_opt *opt, const char *name, char *buf,
 		from = opt->last_shown + 1;
 
 	/* Rewind. */
-	while (bol > buf &&
+	while (bol > gs->buf &&
 	       cur > (funcname_needed == 2 ? opt->last_shown + 1 : from)) {
 		char *eol = --bol;
 
-		while (bol > buf && bol[-1] != '\n')
+		while (bol > gs->buf && bol[-1] != '\n')
 			bol--;
 		cur--;
-		if (funcname_needed && match_funcname(opt, name, bol, eol)) {
+		if (funcname_needed && match_funcname(opt, gs, bol, eol)) {
 			funcname_lno = cur;
 			funcname_needed = 0;
 		}
@@ -916,7 +916,7 @@ static void show_pre_context(struct grep_opt *opt, const char *name, char *buf,
 
 	/* We need to look even further back to find a function signature. */
 	if (opt->funcname && funcname_needed)
-		show_funcname_line(opt, name, buf, bol, cur);
+		show_funcname_line(opt, gs, bol, cur);
 
 	/* Back forward. */
 	while (cur < lno) {
@@ -924,7 +924,7 @@ static void show_pre_context(struct grep_opt *opt, const char *name, char *buf,
 
 		while (*eol != '\n')
 			eol++;
-		show_line(opt, bol, eol, name, cur, sign);
+		show_line(opt, bol, eol, gs->name, cur, sign);
 		bol = eol + 1;
 		cur++;
 	}
@@ -991,11 +991,10 @@ static void std_output(struct grep_opt *opt, const void *buf, size_t size)
 	fwrite(buf, size, 1, stdout);
 }
 
-static int grep_buffer_1(struct grep_opt *opt, const char *name,
-			 char *buf, unsigned long size, int collect_hits)
+static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int collect_hits)
 {
-	char *bol = buf;
-	unsigned long left = size;
+	char *bol;
+	unsigned long left;
 	unsigned lno = 1;
 	unsigned last_hit = 0;
 	int binary_match_only = 0;
@@ -1023,13 +1022,16 @@ static int grep_buffer_1(struct grep_opt *opt, const char *name,
 	}
 	opt->last_shown = 0;
 
+	if (grep_source_load(gs) < 0)
+		return 0;
+
 	switch (opt->binary) {
 	case GREP_BINARY_DEFAULT:
-		if (buffer_is_binary(buf, size))
+		if (buffer_is_binary(gs->buf, gs->size))
 			binary_match_only = 1;
 		break;
 	case GREP_BINARY_NOMATCH:
-		if (buffer_is_binary(buf, size))
+		if (buffer_is_binary(gs->buf, gs->size))
 			return 0; /* Assume unmatch */
 		break;
 	case GREP_BINARY_TEXT:
@@ -1043,6 +1045,8 @@ static int grep_buffer_1(struct grep_opt *opt, const char *name,
 
 	try_lookahead = should_lookahead(opt);
 
+	bol = gs->buf;
+	left = gs->size;
 	while (left) {
 		char *eol, ch;
 		int hit;
@@ -1091,14 +1095,14 @@ static int grep_buffer_1(struct grep_opt *opt, const char *name,
 			if (opt->status_only)
 				return 1;
 			if (opt->name_only) {
-				show_name(opt, name);
+				show_name(opt, gs->name);
 				return 1;
 			}
 			if (opt->count)
 				goto next_line;
 			if (binary_match_only) {
 				opt->output(opt, "Binary file ", 12);
-				output_color(opt, name, strlen(name),
+				output_color(opt, gs->name, strlen(gs->name),
 					     opt->color_filename);
 				opt->output(opt, " matches\n", 9);
 				return 1;
@@ -1107,23 +1111,23 @@ static int grep_buffer_1(struct grep_opt *opt, const char *name,
 			 * pre-context lines, we would need to show them.
 			 */
 			if (opt->pre_context || opt->funcbody)
-				show_pre_context(opt, name, buf, bol, eol, lno);
+				show_pre_context(opt, gs, bol, eol, lno);
 			else if (opt->funcname)
-				show_funcname_line(opt, name, buf, bol, lno);
-			show_line(opt, bol, eol, name, lno, ':');
+				show_funcname_line(opt, gs, bol, lno);
+			show_line(opt, bol, eol, gs->name, lno, ':');
 			last_hit = lno;
 			if (opt->funcbody)
 				show_function = 1;
 			goto next_line;
 		}
-		if (show_function && match_funcname(opt, name, bol, eol))
+		if (show_function && match_funcname(opt, gs, bol, eol))
 			show_function = 0;
 		if (show_function ||
 		    (last_hit && lno <= last_hit + opt->post_context)) {
 			/* If the last hit is within the post context,
 			 * we need to show this line.
 			 */
-			show_line(opt, bol, eol, name, lno, '-');
+			show_line(opt, bol, eol, gs->name, lno, '-');
 		}
 
 	next_line:
@@ -1141,7 +1145,7 @@ static int grep_buffer_1(struct grep_opt *opt, const char *name,
 		return 0;
 	if (opt->unmatch_name_only) {
 		/* We did not see any hit, so we want to show this */
-		show_name(opt, name);
+		show_name(opt, gs->name);
 		return 1;
 	}
 
@@ -1155,7 +1159,7 @@ static int grep_buffer_1(struct grep_opt *opt, const char *name,
 	 */
 	if (opt->count && count) {
 		char buf[32];
-		output_color(opt, name, strlen(name), opt->color_filename);
+		output_color(opt, gs->name, strlen(gs->name), opt->color_filename);
 		output_sep(opt, ':');
 		snprintf(buf, sizeof(buf), "%u\n", count);
 		opt->output(opt, buf, strlen(buf));
@@ -1190,23 +1194,149 @@ static int chk_hit_marker(struct grep_expr *x)
 	}
 }
 
-int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
+int grep_source(struct grep_opt *opt, struct grep_source *gs)
 {
 	/*
 	 * we do not have to do the two-pass grep when we do not check
 	 * buffer-wide "all-match".
 	 */
 	if (!opt->all_match)
-		return grep_buffer_1(opt, name, buf, size, 0);
+		return grep_source_1(opt, gs, 0);
 
 	/* Otherwise the toplevel "or" terms hit a bit differently.
 	 * We first clear hit markers from them.
 	 */
 	clr_hit_marker(opt->pattern_expression);
-	grep_buffer_1(opt, name, buf, size, 1);
+	grep_source_1(opt, gs, 1);
 
 	if (!chk_hit_marker(opt->pattern_expression))
 		return 0;
 
-	return grep_buffer_1(opt, name, buf, size, 0);
+	return grep_source_1(opt, gs, 0);
+}
+
+int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
+{
+	struct grep_source gs;
+	int r;
+
+	grep_source_init(&gs, GREP_SOURCE_BUF, name, NULL);
+	gs.buf = buf;
+	gs.size = size;
+
+	r = grep_source(opt, &gs);
+
+	grep_source_clear(&gs);
+	return r;
+}
+
+void grep_source_init(struct grep_source *gs, enum grep_source_type type,
+		      const char *name, const void *identifier)
+{
+	gs->type = type;
+	gs->name = name ? xstrdup(name) : NULL;
+	gs->buf = NULL;
+	gs->size = 0;
+
+	switch (type) {
+	case GREP_SOURCE_FILE:
+		gs->identifier = xstrdup(identifier);
+		break;
+	case GREP_SOURCE_SHA1:
+		gs->identifier = xmalloc(20);
+		memcpy(gs->identifier, identifier, 20);
+		break;
+	case GREP_SOURCE_BUF:
+		gs->identifier = NULL;
+	}
+}
+
+void grep_source_clear(struct grep_source *gs)
+{
+	free(gs->name);
+	gs->name = NULL;
+	free(gs->identifier);
+	gs->identifier = NULL;
+	grep_source_clear_data(gs);
+}
+
+void grep_source_clear_data(struct grep_source *gs)
+{
+	switch (gs->type) {
+	case GREP_SOURCE_FILE:
+	case GREP_SOURCE_SHA1:
+		free(gs->buf);
+		gs->buf = NULL;
+		gs->size = 0;
+		break;
+	case GREP_SOURCE_BUF:
+		/* leave user-provided buf intact */
+		break;
+	}
+}
+
+static int grep_source_load_sha1(struct grep_source *gs)
+{
+	enum object_type type;
+
+	grep_read_lock();
+	gs->buf = read_sha1_file(gs->identifier, &type, &gs->size);
+	grep_read_unlock();
+
+	if (!gs->buf)
+		return error(_("'%s': unable to read %s"),
+			     gs->name,
+			     sha1_to_hex(gs->identifier));
+	return 0;
+}
+
+static int grep_source_load_file(struct grep_source *gs)
+{
+	const char *filename = gs->identifier;
+	struct stat st;
+	char *data;
+	size_t size;
+	int i;
+
+	if (lstat(filename, &st) < 0) {
+	err_ret:
+		if (errno != ENOENT)
+			error(_("'%s': %s"), filename, strerror(errno));
+		return -1;
+	}
+	if (!S_ISREG(st.st_mode))
+		return -1;
+	size = xsize_t(st.st_size);
+	i = open(filename, O_RDONLY);
+	if (i < 0)
+		goto err_ret;
+	data = xmalloc(size + 1);
+	if (st.st_size != read_in_full(i, data, size)) {
+		error(_("'%s': short read %s"), filename, strerror(errno));
+		close(i);
+		free(data);
+		return -1;
+	}
+	close(i);
+	data[size] = 0;
+
+	gs->buf = data;
+	gs->size = size;
+	return 0;
+}
+
+int grep_source_load(struct grep_source *gs)
+{
+	if (gs->buf)
+		return 0;
+
+	switch (gs->type) {
+	case GREP_SOURCE_FILE:
+		return grep_source_load_file(gs);
+	case GREP_SOURCE_SHA1:
+		return grep_source_load_sha1(gs);
+	case GREP_SOURCE_BUF:
+		return gs->buf ? 0 : -1;
+	}
+	die("BUG: invalid grep_source type");
 }
diff --git a/grep.h b/grep.h
index 4f1b025..e386ca4 100644
--- a/grep.h
+++ b/grep.h
@@ -129,6 +129,28 @@ extern void compile_grep_patterns(struct grep_opt *opt);
 extern void free_grep_patterns(struct grep_opt *opt);
 extern int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size);
 
+struct grep_source {
+	char *name;
+
+	enum grep_source_type {
+		GREP_SOURCE_SHA1,
+		GREP_SOURCE_FILE,
+		GREP_SOURCE_BUF,
+	} type;
+	void *identifier;
+
+	char *buf;
+	unsigned long size;
+};
+
+void grep_source_init(struct grep_source *gs, enum grep_source_type type,
+		      const char *name, const void *identifier);
+int grep_source_load(struct grep_source *gs);
+void grep_source_clear_data(struct grep_source *gs);
+void grep_source_clear(struct grep_source *gs);
+
+int grep_source(struct grep_opt *opt, struct grep_source *gs);
+
 extern struct grep_opt *grep_opt_dup(const struct grep_opt *opt);
 extern int grep_threads_ok(const struct grep_opt *opt);
 
-- 
cgit v0.10.2-6-g49f6


From 8f24a6323ece9be1bf1a04b4b5856112438337f2 Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:19:37 -0500
Subject: convert git-grep to use grep_source interface

The grep_source interface (as opposed to grep_buffer) will
eventually gives us a richer interface for telling the
low-level grep code about our buffers. Eventually this will
lead to things like better binary-file handling. For now, it
lets us drop a lot of now-redundant code.

The conversion is mostly straight-forward. One thing to note
is that the memory ownership rules for "struct grep_source"
are different than the "struct work_item" found here (the
former will copy things like the filename, rather than
taking ownership). Therefore you will also see some slight
tweaking of when filename buffers are released.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/builtin/grep.c b/builtin/grep.c
index aeb3616..c9b6a13 100644
--- a/builtin/grep.c
+++ b/builtin/grep.c
@@ -29,25 +29,12 @@ static int use_threads = 1;
 #define THREADS 8
 static pthread_t threads[THREADS];
 
-static void *load_sha1(const unsigned char *sha1, unsigned long *size,
-		       const char *name);
-static void *load_file(const char *filename, size_t *sz);
-
-enum work_type {WORK_SHA1, WORK_FILE};
-
 /* We use one producer thread and THREADS consumer
  * threads. The producer adds struct work_items to 'todo' and the
  * consumers pick work items from the same array.
  */
 struct work_item {
-	enum work_type type;
-	char *name;
-
-	/* if type == WORK_SHA1, then 'identifier' is a SHA1,
-	 * otherwise type == WORK_FILE, and 'identifier' is a NUL
-	 * terminated filename.
-	 */
-	void *identifier;
+	struct grep_source source;
 	char done;
 	struct strbuf out;
 };
@@ -98,7 +85,8 @@ static pthread_cond_t cond_result;
 
 static int skip_first_line;
 
-static void add_work(enum work_type type, char *name, void *id)
+static void add_work(enum grep_source_type type, const char *name,
+		     const void *id)
 {
 	grep_lock();
 
@@ -106,9 +94,7 @@ static void add_work(enum work_type type, char *name, void *id)
 		pthread_cond_wait(&cond_write, &grep_mutex);
 	}
 
-	todo[todo_end].type = type;
-	todo[todo_end].name = name;
-	todo[todo_end].identifier = id;
+	grep_source_init(&todo[todo_end].source, type, name, id);
 	todo[todo_end].done = 0;
 	strbuf_reset(&todo[todo_end].out);
 	todo_end = (todo_end + 1) % ARRAY_SIZE(todo);
@@ -136,21 +122,6 @@ static struct work_item *get_work(void)
 	return ret;
 }
 
-static void grep_sha1_async(struct grep_opt *opt, char *name,
-			    const unsigned char *sha1)
-{
-	unsigned char *s;
-	s = xmalloc(20);
-	memcpy(s, sha1, 20);
-	add_work(WORK_SHA1, name, s);
-}
-
-static void grep_file_async(struct grep_opt *opt, char *name,
-			    const char *filename)
-{
-	add_work(WORK_FILE, name, xstrdup(filename));
-}
-
 static void work_done(struct work_item *w)
 {
 	int old_done;
@@ -177,8 +148,7 @@ static void work_done(struct work_item *w)
 
 			write_or_die(1, p, len);
 		}
-		free(w->name);
-		free(w->identifier);
+		grep_source_clear(&w->source);
 	}
 
 	if (old_done != todo_done)
@@ -201,25 +171,8 @@ static void *run(void *arg)
 			break;
 
 		opt->output_priv = w;
-		if (w->type == WORK_SHA1) {
-			unsigned long sz;
-			void* data = load_sha1(w->identifier, &sz, w->name);
-
-			if (data) {
-				hit |= grep_buffer(opt, w->name, data, sz);
-				free(data);
-			}
-		} else if (w->type == WORK_FILE) {
-			size_t sz;
-			void* data = load_file(w->identifier, &sz);
-			if (data) {
-				hit |= grep_buffer(opt, w->name, data, sz);
-				free(data);
-			}
-		} else {
-			assert(0);
-		}
-
+		hit |= grep_source(opt, &w->source);
+		grep_source_clear_data(&w->source);
 		work_done(w);
 	}
 	free_grep_patterns(arg);
@@ -365,23 +318,10 @@ static void *lock_and_read_sha1_file(const unsigned char *sha1, enum object_type
 	return data;
 }
 
-static void *load_sha1(const unsigned char *sha1, unsigned long *size,
-		       const char *name)
-{
-	enum object_type type;
-	void *data = lock_and_read_sha1_file(sha1, &type, size);
-
-	if (!data)
-		error(_("'%s': unable to read %s"), name, sha1_to_hex(sha1));
-
-	return data;
-}
-
 static int grep_sha1(struct grep_opt *opt, const unsigned char *sha1,
 		     const char *filename, int tree_name_len)
 {
 	struct strbuf pathbuf = STRBUF_INIT;
-	char *name;
 
 	if (opt->relative && opt->prefix_length) {
 		quote_path_relative(filename + tree_name_len, -1, &pathbuf,
@@ -391,87 +331,51 @@ static int grep_sha1(struct grep_opt *opt, const unsigned char *sha1,
 		strbuf_addstr(&pathbuf, filename);
 	}
 
-	name = strbuf_detach(&pathbuf, NULL);
-
 #ifndef NO_PTHREADS
 	if (use_threads) {
-		grep_sha1_async(opt, name, sha1);
+		add_work(GREP_SOURCE_SHA1, pathbuf.buf, sha1);
+		strbuf_release(&pathbuf);
 		return 0;
 	} else
 #endif
 	{
+		struct grep_source gs;
 		int hit;
-		unsigned long sz;
-		void *data = load_sha1(sha1, &sz, name);
-		if (!data)
-			hit = 0;
-		else
-			hit = grep_buffer(opt, name, data, sz);
 
-		free(data);
-		free(name);
-		return hit;
-	}
-}
+		grep_source_init(&gs, GREP_SOURCE_SHA1, pathbuf.buf, sha1);
+		strbuf_release(&pathbuf);
+		hit = grep_source(opt, &gs);
 
-static void *load_file(const char *filename, size_t *sz)
-{
-	struct stat st;
-	char *data;
-	int i;
-
-	if (lstat(filename, &st) < 0) {
-	err_ret:
-		if (errno != ENOENT)
-			error(_("'%s': %s"), filename, strerror(errno));
-		return NULL;
-	}
-	if (!S_ISREG(st.st_mode))
-		return NULL;
-	*sz = xsize_t(st.st_size);
-	i = open(filename, O_RDONLY);
-	if (i < 0)
-		goto err_ret;
-	data = xmalloc(*sz + 1);
-	if (st.st_size != read_in_full(i, data, *sz)) {
-		error(_("'%s': short read %s"), filename, strerror(errno));
-		close(i);
-		free(data);
-		return NULL;
+		grep_source_clear(&gs);
+		return hit;
 	}
-	close(i);
-	data[*sz] = 0;
-	return data;
 }
 
 static int grep_file(struct grep_opt *opt, const char *filename)
 {
 	struct strbuf buf = STRBUF_INIT;
-	char *name;
 
 	if (opt->relative && opt->prefix_length)
 		quote_path_relative(filename, -1, &buf, opt->prefix);
 	else
 		strbuf_addstr(&buf, filename);
-	name = strbuf_detach(&buf, NULL);
 
 #ifndef NO_PTHREADS
 	if (use_threads) {
-		grep_file_async(opt, name, filename);
+		add_work(GREP_SOURCE_FILE, buf.buf, filename);
+		strbuf_release(&buf);
 		return 0;
 	} else
 #endif
 	{
+		struct grep_source gs;
 		int hit;
-		size_t sz;
-		void *data = load_file(filename, &sz);
-		if (!data)
-			hit = 0;
-		else
-			hit = grep_buffer(opt, name, data, sz);
 
-		free(data);
-		free(name);
+		grep_source_init(&gs, GREP_SOURCE_FILE, buf.buf, filename);
+		strbuf_release(&buf);
+		hit = grep_source(opt, &gs);
+
+		grep_source_clear(&gs);
 		return hit;
 	}
 }
-- 
cgit v0.10.2-6-g49f6


From c876d6da88d9bf1f3377d4eed66fc7705e31c30e Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:20:10 -0500
Subject: grep: drop grep_buffer's "name" parameter

Before the grep_source interface existed, grep_buffer was
used by two types of callers:

  1. Ones which pulled a file into a buffer, and then wanted
     to supply the file's name for the output (i.e.,
     git grep).

  2. Ones which really just wanted to grep a buffer (i.e.,
     git log --grep).

Callers in set (1) should now be using grep_source. Callers
in set (2) always pass NULL for the "name" parameter of
grep_buffer. We can therefore get rid of this now-useless
parameter.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/grep.c b/grep.c
index 8204ca2..2a3fe7c 100644
--- a/grep.c
+++ b/grep.c
@@ -1215,12 +1215,12 @@ int grep_source(struct grep_opt *opt, struct grep_source *gs)
 	return grep_source_1(opt, gs, 0);
 }
 
-int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size)
+int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size)
 {
 	struct grep_source gs;
 	int r;
 
-	grep_source_init(&gs, GREP_SOURCE_BUF, name, NULL);
+	grep_source_init(&gs, GREP_SOURCE_BUF, NULL, NULL);
 	gs.buf = buf;
 	gs.size = size;
 
diff --git a/grep.h b/grep.h
index e386ca4..8bf3001 100644
--- a/grep.h
+++ b/grep.h
@@ -127,7 +127,7 @@ extern void append_grep_pattern(struct grep_opt *opt, const char *pat, const cha
 extern void append_header_grep_pattern(struct grep_opt *, enum grep_header_field, const char *);
 extern void compile_grep_patterns(struct grep_opt *opt);
 extern void free_grep_patterns(struct grep_opt *opt);
-extern int grep_buffer(struct grep_opt *opt, const char *name, char *buf, unsigned long size);
+extern int grep_buffer(struct grep_opt *opt, char *buf, unsigned long size);
 
 struct grep_source {
 	char *name;
diff --git a/revision.c b/revision.c
index 064e351..89d3a8f 100644
--- a/revision.c
+++ b/revision.c
@@ -2128,7 +2128,6 @@ static int commit_match(struct commit *commit, struct rev_info *opt)
 	if (!opt->grep_filter.pattern_list && !opt->grep_filter.header_list)
 		return 1;
 	return grep_buffer(&opt->grep_filter,
-			   NULL, /* we say nothing, not even filename */
 			   commit->buffer, strlen(commit->buffer));
 }
 
-- 
cgit v0.10.2-6-g49f6


From 94ad9d9e075d3f7802cf56641ebb342d43fb46e3 Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:20:43 -0500
Subject: grep: cache userdiff_driver in grep_source

Right now, grep only uses the userdiff_driver for one thing:
looking up funcname patterns for "-p" and "-W".  As new uses
for userdiff drivers are added to the grep code, we want to
minimize attribute lookups, which can be expensive.

It might seem at first that this would also optimize multiple
lookups when the funcname pattern for a file is needed
multiple times. However, the compiled funcname pattern is
already cached in struct grep_opt's "priv" member, so
multiple lookups are already suppressed.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/grep.c b/grep.c
index 2a3fe7c..bb18569 100644
--- a/grep.c
+++ b/grep.c
@@ -841,12 +841,9 @@ static int match_funcname(struct grep_opt *opt, struct grep_source *gs, char *bo
 {
 	xdemitconf_t *xecfg = opt->priv;
 	if (xecfg && !xecfg->find_func) {
-		struct userdiff_driver *drv;
-		grep_attr_lock();
-		drv = userdiff_find_by_path(gs->name);
-		grep_attr_unlock();
-		if (drv && drv->funcname.pattern) {
-			const struct userdiff_funcname *pe = &drv->funcname;
+		grep_source_load_driver(gs);
+		if (gs->driver->funcname.pattern) {
+			const struct userdiff_funcname *pe = &gs->driver->funcname;
 			xdiff_set_find_func(xecfg, pe->pattern, pe->cflags);
 		} else {
 			xecfg = opt->priv = NULL;
@@ -1237,6 +1234,7 @@ void grep_source_init(struct grep_source *gs, enum grep_source_type type,
 	gs->name = name ? xstrdup(name) : NULL;
 	gs->buf = NULL;
 	gs->size = 0;
+	gs->driver = NULL;
 
 	switch (type) {
 	case GREP_SOURCE_FILE:
@@ -1340,3 +1338,15 @@ int grep_source_load(struct grep_source *gs)
 	}
 	die("BUG: invalid grep_source type");
 }
+
+void grep_source_load_driver(struct grep_source *gs)
+{
+	if (gs->driver)
+		return;
+
+	grep_attr_lock();
+	gs->driver = userdiff_find_by_path(gs->name);
+	if (!gs->driver)
+		gs->driver = userdiff_find_by_name("default");
+	grep_attr_unlock();
+}
diff --git a/grep.h b/grep.h
index 8bf3001..73b28c2 100644
--- a/grep.h
+++ b/grep.h
@@ -9,6 +9,7 @@ typedef int pcre_extra;
 #endif
 #include "kwset.h"
 #include "thread-utils.h"
+#include "userdiff.h"
 
 enum grep_pat_token {
 	GREP_PATTERN,
@@ -141,6 +142,8 @@ struct grep_source {
 
 	char *buf;
 	unsigned long size;
+
+	struct userdiff_driver *driver;
 };
 
 void grep_source_init(struct grep_source *gs, enum grep_source_type type,
@@ -148,6 +151,7 @@ void grep_source_init(struct grep_source *gs, enum grep_source_type type,
 int grep_source_load(struct grep_source *gs);
 void grep_source_clear_data(struct grep_source *gs);
 void grep_source_clear(struct grep_source *gs);
+void grep_source_load_driver(struct grep_source *gs);
 
 int grep_source(struct grep_opt *opt, struct grep_source *gs);
 
-- 
cgit v0.10.2-6-g49f6


From 41b59bfcb16abb738e5c95c95fb462e717d47d4d Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:21:02 -0500
Subject: grep: respect diff attributes for binary-ness

There is currently no way for users to tell git-grep that a
particular path is or is not a binary file; instead, grep
always relies on its auto-detection (or the user specifying
"-a" to treat all binary-looking files like text).

This patch teaches git-grep to use the same attribute lookup
that is used by git-diff. We could add a new "grep" flag,
but that is unnecessarily complex and unlikely to be useful.
Despite the name, the "-diff" attribute (or "diff=foo" and
the associated diff.foo.binary config option) are really
about describing the contents of the path. It's simply
historical that diff was the only thing that cared about
these attributes in the past.

And if this simple approach turns out to be insufficient, we
still have a backwards-compatible path forward: we can add a
separate "grep" attribute, and fall back to respecting
"diff" if it is unset.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/grep.c b/grep.c
index bb18569..a50d161 100644
--- a/grep.c
+++ b/grep.c
@@ -1024,11 +1024,11 @@ static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int colle
 
 	switch (opt->binary) {
 	case GREP_BINARY_DEFAULT:
-		if (buffer_is_binary(gs->buf, gs->size))
+		if (grep_source_is_binary(gs))
 			binary_match_only = 1;
 		break;
 	case GREP_BINARY_NOMATCH:
-		if (buffer_is_binary(gs->buf, gs->size))
+		if (grep_source_is_binary(gs))
 			return 0; /* Assume unmatch */
 		break;
 	case GREP_BINARY_TEXT:
@@ -1350,3 +1350,15 @@ void grep_source_load_driver(struct grep_source *gs)
 		gs->driver = userdiff_find_by_name("default");
 	grep_attr_unlock();
 }
+
+int grep_source_is_binary(struct grep_source *gs)
+{
+	grep_source_load_driver(gs);
+	if (gs->driver->binary != -1)
+		return gs->driver->binary;
+
+	if (!grep_source_load(gs))
+		return buffer_is_binary(gs->buf, gs->size);
+
+	return 0;
+}
diff --git a/grep.h b/grep.h
index 73b28c2..36e49d8 100644
--- a/grep.h
+++ b/grep.h
@@ -152,6 +152,7 @@ int grep_source_load(struct grep_source *gs);
 void grep_source_clear_data(struct grep_source *gs);
 void grep_source_clear(struct grep_source *gs);
 void grep_source_load_driver(struct grep_source *gs);
+int grep_source_is_binary(struct grep_source *gs);
 
 int grep_source(struct grep_opt *opt, struct grep_source *gs);
 
diff --git a/t/t7008-grep-binary.sh b/t/t7008-grep-binary.sh
index 917a264..fd6410f 100755
--- a/t/t7008-grep-binary.sh
+++ b/t/t7008-grep-binary.sh
@@ -99,4 +99,28 @@ test_expect_success 'git grep y<NUL>x a' "
 	test_must_fail git grep -f f a
 "
 
+test_expect_success 'grep respects binary diff attribute' '
+	echo text >t &&
+	git add t &&
+	echo t:text >expect &&
+	git grep text t >actual &&
+	test_cmp expect actual &&
+	echo "t -diff" >.gitattributes &&
+	echo "Binary file t matches" >expect &&
+	git grep text t >actual &&
+	test_cmp expect actual
+'
+
+test_expect_success 'grep respects not-binary diff attribute' '
+	echo binQary | q_to_nul >b &&
+	git add b &&
+	echo "Binary file b matches" >expect &&
+	git grep bin b >actual &&
+	test_cmp expect actual &&
+	echo "b diff" >.gitattributes &&
+	echo "b:binQary" >expect &&
+	git grep bin b | nul_to_q >actual &&
+	test_cmp expect actual
+'
+
 test_done
-- 
cgit v0.10.2-6-g49f6


From 08265798e1ff6abc1b0aaff31c1471f83bd51425 Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:21:11 -0500
Subject: grep: load file data after checking binary-ness

Usually we load each file to grep into memory, check whether
it's binary, and then either grep it (the default) or not
(if "-I" was given).

In the "-I" case, we can skip loading the file entirely if
it is marked as binary via gitattributes. On my giant
3-gigabyte media repository, doing "git grep -I foo" went
from:

  real    0m0.712s
  user    0m0.044s
  sys     0m4.780s

to:

  real    0m0.026s
  user    0m0.016s
  sys     0m0.020s

Obviously this is an extreme example. The repo is almost
entirely binary files, and you can see that we spent all of
our time asking the kernel to read() the data. However, with
a cold disk cache, even avoiding a few binary files can have
an impact.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/grep.c b/grep.c
index a50d161..3821400 100644
--- a/grep.c
+++ b/grep.c
@@ -1019,9 +1019,6 @@ static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int colle
 	}
 	opt->last_shown = 0;
 
-	if (grep_source_load(gs) < 0)
-		return 0;
-
 	switch (opt->binary) {
 	case GREP_BINARY_DEFAULT:
 		if (grep_source_is_binary(gs))
@@ -1042,6 +1039,9 @@ static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int colle
 
 	try_lookahead = should_lookahead(opt);
 
+	if (grep_source_load(gs) < 0)
+		return 0;
+
 	bol = gs->buf;
 	left = gs->size;
 	while (left) {
-- 
cgit v0.10.2-6-g49f6


From 9dd5245c1043dd18fd7b3f44b9e51eef7e4b58d8 Mon Sep 17 00:00:00 2001
From: Jeff King <peff@peff.net>
Date: Thu, 2 Feb 2012 03:24:28 -0500
Subject: grep: pre-load userdiff drivers when threaded

The low-level grep_source code will automatically load the
userdiff driver to see whether a file is binary. However,
when we are threaded, it will load the drivers in a
non-deterministic order, handling each one as its assigned
thread happens to be scheduled.

Meanwhile, the attribute lookup code (which underlies the
userdiff driver lookup) is optimized to handle paths in
sequential order (because they tend to share the same
gitattributes files). Multi-threading the lookups destroys
the locality and makes this optimization less effective.

We can fix this by pre-loading the userdiff driver in the
main thread, before we hand off the file to a worker thread.
My best-of-five for "git grep foo" on the linux-2.6
repository went from:

  real    0m0.391s
  user    0m1.708s
  sys     0m0.584s

to:

  real    0m0.360s
  user    0m1.576s
  sys     0m0.572s

Not a huge speedup, but it's quite easy to do. The only
trick is that we shouldn't perform this optimization if "-a"
was used, in which case we won't bother checking whether
the files are binary at all.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/builtin/grep.c b/builtin/grep.c
index c9b6a13..e741aca 100644
--- a/builtin/grep.c
+++ b/builtin/grep.c
@@ -85,8 +85,8 @@ static pthread_cond_t cond_result;
 
 static int skip_first_line;
 
-static void add_work(enum grep_source_type type, const char *name,
-		     const void *id)
+static void add_work(struct grep_opt *opt, enum grep_source_type type,
+		     const char *name, const void *id)
 {
 	grep_lock();
 
@@ -95,6 +95,8 @@ static void add_work(enum grep_source_type type, const char *name,
 	}
 
 	grep_source_init(&todo[todo_end].source, type, name, id);
+	if (opt->binary != GREP_BINARY_TEXT)
+		grep_source_load_driver(&todo[todo_end].source);
 	todo[todo_end].done = 0;
 	strbuf_reset(&todo[todo_end].out);
 	todo_end = (todo_end + 1) % ARRAY_SIZE(todo);
@@ -333,7 +335,7 @@ static int grep_sha1(struct grep_opt *opt, const unsigned char *sha1,
 
 #ifndef NO_PTHREADS
 	if (use_threads) {
-		add_work(GREP_SOURCE_SHA1, pathbuf.buf, sha1);
+		add_work(opt, GREP_SOURCE_SHA1, pathbuf.buf, sha1);
 		strbuf_release(&pathbuf);
 		return 0;
 	} else
@@ -362,7 +364,7 @@ static int grep_file(struct grep_opt *opt, const char *filename)
 
 #ifndef NO_PTHREADS
 	if (use_threads) {
-		add_work(GREP_SOURCE_FILE, buf.buf, filename);
+		add_work(opt, GREP_SOURCE_FILE, buf.buf, filename);
 		strbuf_release(&buf);
 		return 0;
 	} else
-- 
cgit v0.10.2-6-g49f6