Log
[[[
Fix delta composer reading past EOF by making it stop reading after the first
partial read.
* libsvn_delta/text_delta.c
(svn_txdelta_stream_t): Add a more_source field to indicate if we can read
more from the source stream.
(svn_txdelta_next_window): Stop reading from stream->source after a partial
read has signalled EOF. Also, use the now-common SVN_ERR() macro to report
errors to callers.
]]]
Index: subversion/libsvn_delta/text_delta.c
===================================================================
--- subversion/libsvn_delta/text_delta.c (revision 17742)
+++ subversion/libsvn_delta/text_delta.c (working copy)
@@ -37,6 +37,7 @@
svn_stream_t *target;
/* Private data */
+ svn_boolean_t more_source; /* FALSE if EOF for the source stream. */
svn_boolean_t more; /* TRUE if there are more data in the pool. */
svn_filesize_t pos; /* Offset of next read in source file. */
char *buf; /* Buffer for vdelta data. */
@@ -272,8 +273,9 @@
apr_pool_t *pool)
{
*stream = apr_palloc (pool, sizeof (**stream));
- (*stream)->source = source;
+ (*stream)->source = source;
(*stream)->target = target;
+ (*stream)->more_source = TRUE;
(*stream)->more = TRUE;
(*stream)->pos = 0;
(*stream)->buf = apr_palloc (pool, 2 * SVN_DELTA_WINDOW_SIZE);
@@ -289,21 +291,23 @@
svn_txdelta_stream_t *stream,
apr_pool_t *pool)
{
- svn_error_t *err;
apr_size_t source_len = SVN_DELTA_WINDOW_SIZE;
apr_size_t target_len = SVN_DELTA_WINDOW_SIZE;
-
+
/* Read the source stream. */
- err = svn_stream_read (stream->source, stream->buf, &source_len);
-
+ if (stream->more_source)
+ {
+ SVN_ERR (svn_stream_read (stream->source, stream->buf, &source_len));
+ stream->more_source = source_len == SVN_DELTA_WINDOW_SIZE;
+ }
+ else
+ source_len = 0;
+
/* Read the target stream. */
- if (err == SVN_NO_ERROR)
- err = svn_stream_read (stream->target, stream->buf + source_len,
- &target_len);
- if (err != SVN_NO_ERROR)
- return err;
+ SVN_ERR (svn_stream_read (stream->target, stream->buf + source_len,
+ &target_len));
stream->pos += source_len;
-
+
/* ### The apr_md5 functions always return APR_SUCCESS. At one
point, we proposed to APR folks that the interfaces change to
return void, but for some people that was apparently not a good