Commit 66633e836f192054746941e274e9e9008ee5cdb3

Carlos Martín Nieto 2016-04-27T12:00:31

odb: avoid inflating the full delta to read the header When we read the header, we want to know the size and type of the object. We're currently inflating the full delta in order to read the first few bytes. This can mean hundreds of kB needlessly inflated for large objects. Instead use a packfile stream to read just enough so we can read the two varints in the header and avoid inflating most of the delta.

diff --git a/src/delta-apply.c b/src/delta-apply.c
index 89745fa..6e86a81 100644
--- a/src/delta-apply.c
+++ b/src/delta-apply.c
@@ -49,6 +49,37 @@ int git__delta_read_header(
 	return 0;
 }
 
+#define DELTA_HEADER_BUFFER_LEN 16
+int git__delta_read_header_fromstream(size_t *base_sz, size_t *res_sz, git_packfile_stream *stream)
+{
+	static const size_t buffer_len = DELTA_HEADER_BUFFER_LEN;
+	unsigned char buffer[DELTA_HEADER_BUFFER_LEN];
+	const unsigned char *delta, *delta_end;
+	size_t len;
+	ssize_t read;
+
+	len = read = 0;
+	while (len < buffer_len) {
+		read = git_packfile_stream_read(stream, &buffer[len], buffer_len - len);
+
+		if (read == 0)
+			break;
+
+		if (read == GIT_EBUFS)
+			continue;
+
+		len += read;
+	}
+
+	delta = buffer;
+	delta_end = delta + len;
+	if ((hdr_sz(base_sz, &delta, delta_end) < 0) ||
+	    (hdr_sz(res_sz, &delta, delta_end) < 0))
+		return -1;
+
+	return 0;
+}
+
 int git__delta_apply(
 	git_rawobj *out,
 	const unsigned char *base,
diff --git a/src/delta-apply.h b/src/delta-apply.h
index d7d99d0..eeeb786 100644
--- a/src/delta-apply.h
+++ b/src/delta-apply.h
@@ -8,6 +8,7 @@
 #define INCLUDE_delta_apply_h__
 
 #include "odb.h"
+#include "pack.h"
 
 /**
  * Apply a git binary delta to recover the original content.
@@ -47,4 +48,15 @@ extern int git__delta_read_header(
 	size_t *base_sz,
 	size_t *res_sz);
 
+/**
+ * Read the header of a git binary delta
+ *
+ * This variant reads just enough from the packfile stream to read the
+ * delta header.
+ */
+extern int git__delta_read_header_fromstream(
+	size_t *base_sz,
+	size_t *res_sz,
+	git_packfile_stream *stream);
+
 #endif
diff --git a/src/pack.c b/src/pack.c
index e7003e6..6a700e2 100644
--- a/src/pack.c
+++ b/src/pack.c
@@ -499,15 +499,14 @@ int git_packfile_resolve_header(
 
 	if (type == GIT_OBJ_OFS_DELTA || type == GIT_OBJ_REF_DELTA) {
 		size_t base_size;
-		git_rawobj delta;
+		git_packfile_stream stream;
+
 		base_offset = get_delta_base(p, &w_curs, &curpos, type, offset);
 		git_mwindow_close(&w_curs);
-		error = packfile_unpack_compressed(&delta, p, &w_curs, &curpos, size, type);
-		git_mwindow_close(&w_curs);
-		if (error < 0)
+		if ((error = git_packfile_stream_open(&stream, p, curpos)) < 0)
 			return error;
-		error = git__delta_read_header(delta.data, delta.len, &base_size, size_p);
-		git__free(delta.data);
+		error = git__delta_read_header_fromstream(&base_size, size_p, &stream);
+		git_packfile_stream_free(&stream);
 		if (error < 0)
 			return error;
 	} else