svn_error_t *
svn_string_from_stream(svn_string_t **result,
svn_stream_t *stream,
apr_pool_t *result_pool,
apr_pool_t *scratch_pool)
{
svn_stringbuf_t *work = svn_stringbuf_create_ensure(SVN__STREAM_CHUNK_SIZE,
result_pool);
char *buffer = apr_palloc(scratch_pool, SVN__STREAM_CHUNK_SIZE);
while (1)
{
apr_size_t len = SVN__STREAM_CHUNK_SIZE;
SVN_ERR(svn_stream_read(stream, buffer, &len));
svn_stringbuf_appendbytes(work, buffer, len);
if (len < SVN__STREAM_CHUNK_SIZE)
break;
}
SVN_ERR(svn_stream_close(stream));
*result = apr_palloc(result_pool, sizeof(**result));
(*result)->data = work->data;
(*result)->len = work->len;
return SVN_NO_ERROR;
}
/* If PATH needs to be escaped, return an escaped version of it, allocated
* from RESULT_POOL. Otherwise, return PATH directly. */
static const char *
auto_escape_path(const char *path,
apr_pool_t *result_pool)
{
apr_size_t len = strlen(path);
apr_size_t i;
const char esc = '\x1b';
for (i = 0; i < len; ++i)
if (path[i] < ' ')
{
svn_stringbuf_t *escaped = svn_stringbuf_create_ensure(2 * len,
result_pool);
for (i = 0; i < len; ++i)
if (path[i] < ' ')
{
svn_stringbuf_appendbyte(escaped, esc);
svn_stringbuf_appendbyte(escaped, path[i] + 'A' - 1);
}
else
{
svn_stringbuf_appendbyte(escaped, path[i]);
}
return escaped->data;
}
return path;
}
/* Generate a string of exactly LEN chars (plus terminating NUL). KEY is
* an arbitrary integer that will be transformed into a character sequence
* using entries of BASIC_STRINGS. The result will be allocated in POOL.
*/
static svn_stringbuf_t *
generate_string(apr_uint64_t key, apr_size_t len, apr_pool_t *pool)
{
svn_stringbuf_t *result = svn_stringbuf_create_ensure(len, pool);
apr_uint64_t temp = key;
apr_uint64_t run = 0;
while (len)
{
apr_size_t idx;
apr_size_t add_len;
if (temp == 0)
{
temp = key;
run++;
}
idx = (temp + run) % STRING_COUNT;
temp /= STRING_COUNT;
add_len = strlen(basic_strings[idx]);
add_len = MIN(len, add_len);
svn_stringbuf_appendbytes(result, basic_strings[idx], add_len);
len -= add_len;
}
return result;
}
/* Read a line of original or modified hunk text from the specified
* RANGE within FILE. FILE is expected to contain unidiff text.
* Leading unidiff symbols ('+', '-', and ' ') are removed from the line,
* Any lines commencing with the VERBOTEN character are discarded.
* VERBOTEN should be '+' or '-', depending on which form of hunk text
* is being read.
*
* All other parameters are as in svn_diff_hunk_readline_original_text()
* and svn_diff_hunk_readline_modified_text().
*/
static svn_error_t *
hunk_readline_original_or_modified(apr_file_t *file,
struct svn_diff__hunk_range *range,
svn_stringbuf_t **stringbuf,
const char **eol,
svn_boolean_t *eof,
char verboten,
apr_pool_t *result_pool,
apr_pool_t *scratch_pool)
{
apr_size_t max_len;
svn_boolean_t filtered;
apr_off_t pos;
svn_stringbuf_t *str;
if (range->current >= range->end)
{
/* We're past the range. Indicate that no bytes can be read. */
*eof = TRUE;
if (eol)
*eol = NULL;
*stringbuf = svn_stringbuf_create("", result_pool);
return SVN_NO_ERROR;
}
pos = 0;
SVN_ERR(svn_io_file_seek(file, APR_CUR, &pos, scratch_pool));
SVN_ERR(svn_io_file_seek(file, APR_SET, &range->current, scratch_pool));
do
{
max_len = range->end - range->current;
SVN_ERR(readline(file, &str, eol, eof, max_len,
result_pool, scratch_pool));
range->current = 0;
SVN_ERR(svn_io_file_seek(file, APR_CUR, &range->current, scratch_pool));
filtered = (str->data[0] == verboten || str->data[0] == '\\');
}
while (filtered && ! *eof);
if (filtered)
{
/* EOF, return an empty string. */
*stringbuf = svn_stringbuf_create_ensure(0, result_pool);
}
else if (str->data[0] == '+' || str->data[0] == '-' || str->data[0] == ' ')
{
/* Shave off leading unidiff symbols. */
*stringbuf = svn_stringbuf_create(str->data + 1, result_pool);
}
else
{
/* Return the line as-is. */
*stringbuf = svn_stringbuf_dup(str, result_pool);
}
SVN_ERR(svn_io_file_seek(file, APR_SET, &pos, scratch_pool));
return SVN_NO_ERROR;
}
/* Helper for svn_repos_dump_fs.
Write a revision record of REV in FS to writable STREAM, using POOL.
*/
static svn_error_t *
write_revision_record(svn_stream_t *stream,
svn_fs_t *fs,
svn_revnum_t rev,
apr_pool_t *pool)
{
apr_size_t len;
apr_hash_t *props;
svn_stringbuf_t *encoded_prophash;
apr_time_t timetemp;
svn_string_t *datevalue;
svn_stream_t *propstream;
/* Read the revision props even if we're aren't going to dump
them for verification purposes */
SVN_ERR(svn_fs_revision_proplist(&props, fs, rev, pool));
/* Run revision date properties through the time conversion to
canonicalize them. */
/* ### Remove this when it is no longer needed for sure. */
datevalue = apr_hash_get(props, SVN_PROP_REVISION_DATE,
APR_HASH_KEY_STRING);
if (datevalue)
{
SVN_ERR(svn_time_from_cstring(&timetemp, datevalue->data, pool));
datevalue = svn_string_create(svn_time_to_cstring(timetemp, pool),
pool);
apr_hash_set(props, SVN_PROP_REVISION_DATE, APR_HASH_KEY_STRING,
datevalue);
}
encoded_prophash = svn_stringbuf_create_ensure(0, pool);
propstream = svn_stream_from_stringbuf(encoded_prophash, pool);
SVN_ERR(svn_hash_write2(props, propstream, "PROPS-END", pool));
SVN_ERR(svn_stream_close(propstream));
/* ### someday write a revision-content-checksum */
SVN_ERR(svn_stream_printf(stream, pool,
SVN_REPOS_DUMPFILE_REVISION_NUMBER
": %ld\n", rev));
SVN_ERR(svn_stream_printf(stream, pool,
SVN_REPOS_DUMPFILE_PROP_CONTENT_LENGTH
": %" APR_SIZE_T_FMT "\n",
encoded_prophash->len));
/* Write out a regular Content-length header for the benefit of
non-Subversion RFC-822 parsers. */
SVN_ERR(svn_stream_printf(stream, pool,
SVN_REPOS_DUMPFILE_CONTENT_LENGTH
": %" APR_SIZE_T_FMT "\n\n",
encoded_prophash->len));
len = encoded_prophash->len;
SVN_ERR(svn_stream_write(stream, encoded_prophash->data, &len));
len = 1;
return svn_stream_write(stream, "\n", &len);
}
svn_stringbuf_t *
svn_skel__unparse(const svn_skel_t *skel, apr_pool_t *pool)
{
svn_stringbuf_t *str
= svn_stringbuf_create_ensure(estimate_unparsed_size(skel) + 200, pool);
return unparse(skel, str);
}
apr_status_t
svn_subr__win32_xlate_to_stringbuf(win32_xlate_t *handle,
const char *src_data,
apr_size_t src_length,
svn_stringbuf_t **dest,
apr_pool_t *pool)
{
WCHAR * wide_str;
int retval, wide_size;
if (src_length == 0)
{
*dest = svn_stringbuf_create_empty(pool);
return APR_SUCCESS;
}
retval = MultiByteToWideChar(handle->from_page_id, 0, src_data, src_length,
NULL, 0);
if (retval == 0)
return apr_get_os_error();
wide_size = retval;
/* Allocate temporary buffer for small strings on stack instead of heap. */
if (wide_size <= MAX_PATH)
{
wide_str = alloca(wide_size * sizeof(WCHAR));
}
else
{
wide_str = apr_palloc(pool, wide_size * sizeof(WCHAR));
}
retval = MultiByteToWideChar(handle->from_page_id, 0, src_data, src_length,
wide_str, wide_size);
if (retval == 0)
return apr_get_os_error();
retval = WideCharToMultiByte(handle->to_page_id, 0, wide_str, wide_size,
NULL, 0, NULL, NULL);
if (retval == 0)
return apr_get_os_error();
/* Ensure that buffer is enough to hold result string and termination
character. */
*dest = svn_stringbuf_create_ensure(retval + 1, pool);
(*dest)->len = retval;
retval = WideCharToMultiByte(handle->to_page_id, 0, wide_str, wide_size,
(*dest)->data, (*dest)->len, NULL, NULL);
if (retval == 0)
return apr_get_os_error();
(*dest)->len = retval;
return APR_SUCCESS;
}
inline Try<std::string> patch(const std::string& s, const Diff& diff)
{
// Initialize the Apache Portable Runtime subsystem, as necessary
// for using the svn library.
initialize();
// Note that svn_pool_create wraps apr_pool_create_ex, which is
// thread safe, see: http://goo.gl/NX0hps.
apr_pool_t* pool = svn_pool_create(NULL);
// We want to apply the svndiff format diff to the source trying to
// produce a result. First setup a handler for applying a text delta
// to the source stream.
svn_string_t source;
source.data = s.data();
source.len = s.length();
svn_txdelta_window_handler_t handler;
void* baton = NULL;
svn_stringbuf_t* patched = svn_stringbuf_create_ensure(s.length(), pool);
svn_txdelta_apply(
svn_stream_from_string(&source, pool),
svn_stream_from_stringbuf(patched, pool),
NULL,
NULL,
pool,
&handler,
&baton);
// Setup a stream that converts an svndiff format diff to a text
// delta, so that we can use our handler to patch the source string.
svn_stream_t* stream = svn_txdelta_parse_svndiff(
handler,
baton,
TRUE,
pool);
// Now feed the diff into the stream to compute the patched result.
const char* data = diff.data.data();
apr_size_t length = diff.data.length();
svn_error_t* error = svn_stream_write(stream, data, &length);
if (error != NULL) {
char buffer[1024];
std::string message(svn_err_best_message(error, buffer, 1024));
svn_pool_destroy(pool);
return Error(message);
}
std::string result(patched->data, patched->len);
svn_pool_destroy(pool);
return result;
}
svn_stringbuf_t *
svn_stringbuf_ncreate(const char *bytes, apr_size_t size, apr_pool_t *pool)
{
svn_stringbuf_t *strbuf = svn_stringbuf_create_ensure(size, pool);
memcpy(strbuf->data, bytes, size);
/* Null termination is the convention -- even if we suspect the data
to be binary, it's not up to us to decide, it's the caller's
call. Heck, that's why they call it the caller! */
strbuf->data[size] = '\0';
strbuf->len = size;
return strbuf;
}
/* Print dumpstream-formatted information about REVISION.
* Implements the `svn_ra_replay_revstart_callback_t' interface.
*/
static svn_error_t *
replay_revstart(svn_revnum_t revision,
void *replay_baton,
const svn_delta_editor_t **editor,
void **edit_baton,
apr_hash_t *rev_props,
apr_pool_t *pool)
{
struct replay_baton *rb = replay_baton;
apr_hash_t *normal_props;
svn_stringbuf_t *propstring;
svn_stream_t *stdout_stream;
svn_stream_t *revprop_stream;
SVN_ERR(svn_stream_for_stdout(&stdout_stream, pool));
/* Revision-number: 19 */
SVN_ERR(svn_stream_printf(stdout_stream, pool,
SVN_REPOS_DUMPFILE_REVISION_NUMBER
": %ld\n", revision));
SVN_ERR(svn_rdump__normalize_props(&normal_props, rev_props, pool));
propstring = svn_stringbuf_create_ensure(0, pool);
revprop_stream = svn_stream_from_stringbuf(propstring, pool);
SVN_ERR(svn_hash_write2(normal_props, revprop_stream, "PROPS-END", pool));
SVN_ERR(svn_stream_close(revprop_stream));
/* Prop-content-length: 13 */
SVN_ERR(svn_stream_printf(stdout_stream, pool,
SVN_REPOS_DUMPFILE_PROP_CONTENT_LENGTH
": %" APR_SIZE_T_FMT "\n", propstring->len));
/* Content-length: 29 */
SVN_ERR(svn_stream_printf(stdout_stream, pool,
SVN_REPOS_DUMPFILE_CONTENT_LENGTH
": %" APR_SIZE_T_FMT "\n\n", propstring->len));
/* Property data. */
SVN_ERR(svn_stream_write(stdout_stream, propstring->data,
&(propstring->len)));
SVN_ERR(svn_stream_printf(stdout_stream, pool, "\n"));
SVN_ERR(svn_stream_close(stdout_stream));
SVN_ERR(svn_rdump__get_dump_editor(editor, edit_baton, revision,
rb->stdout_stream, rb->extra_ra_session,
check_cancel, NULL, pool));
return SVN_NO_ERROR;
}
/* Guts of svn_stream_readline().
* Returns the line read from STREAM in *STRINGBUF, and indicates
* end-of-file in *EOF. If DETECT_EOL is TRUE, the end-of-line indicator
* is detected automatically and returned in *EOL.
* If DETECT_EOL is FALSE, *EOL must point to the desired end-of-line
* indicator. STRINGBUF is allocated in POOL. */
static svn_error_t *
stream_readline_bytewise(svn_stringbuf_t **stringbuf,
svn_boolean_t *eof,
const char *eol,
svn_stream_t *stream,
apr_pool_t *pool)
{
svn_stringbuf_t *str;
apr_size_t numbytes;
const char *match;
char c;
/* Since we're reading one character at a time, let's at least
optimize for the 90% case. 90% of the time, we can avoid the
stringbuf ever having to realloc() itself if we start it out at
80 chars. */
str = svn_stringbuf_create_ensure(LINE_CHUNK_SIZE, pool);
/* Read into STR up to and including the next EOL sequence. */
match = eol;
while (*match)
{
numbytes = 1;
SVN_ERR(svn_stream_read(stream, &c, &numbytes));
if (numbytes != 1)
{
/* a 'short' read means the stream has run out. */
*eof = TRUE;
*stringbuf = str;
return SVN_NO_ERROR;
}
if (c == *match)
match++;
else
match = eol;
svn_stringbuf_appendbyte(str, c);
}
*eof = FALSE;
svn_stringbuf_chop(str, match - eol);
*stringbuf = str;
return SVN_NO_ERROR;
}
/* If the footer data in FILE has not been read, yet, do so now.
* Index locations will only be read upon request as we assume they get
* cached and the FILE is usually used for REP data access only.
* Hence, the separate step.
*/
static svn_error_t *
auto_read_footer(svn_fs_x__revision_file_t *file)
{
if (file->l2p_info.start == -1)
{
apr_off_t filesize = 0;
unsigned char footer_length;
svn_stringbuf_t *footer;
/* Determine file size. */
SVN_ERR(auto_open(file));
SVN_ERR(svn_io_file_seek(file->file, APR_END, &filesize, file->pool));
/* Read last byte (containing the length of the footer). */
SVN_ERR(svn_io_file_aligned_seek(file->file, file->block_size, NULL,
filesize - 1, file->pool));
SVN_ERR(svn_io_file_read_full2(file->file, &footer_length,
sizeof(footer_length), NULL, NULL,
file->pool));
/* Read footer. */
footer = svn_stringbuf_create_ensure(footer_length, file->pool);
SVN_ERR(svn_io_file_aligned_seek(file->file, file->block_size, NULL,
filesize - 1 - footer_length,
file->pool));
SVN_ERR(svn_io_file_read_full2(file->file, footer->data, footer_length,
&footer->len, NULL, file->pool));
footer->data[footer->len] = '\0';
/* Extract index locations. */
SVN_ERR(svn_fs_x__parse_footer(&file->l2p_info.start,
&file->l2p_info.checksum,
&file->p2l_info.start,
&file->p2l_info.checksum,
footer, file->file_info.start_revision,
filesize - footer_length - 1, file->pool));
file->l2p_info.end = file->p2l_info.start;
file->p2l_info.end = filesize - footer_length - 1;
}
return SVN_NO_ERROR;
}
/* If PATH has been escaped, return the un-escaped version of it, allocated
* from RESULT_POOL. Otherwise, return PATH directly. */
static const char *
auto_unescape_path(const char *path,
apr_pool_t *result_pool)
{
const char esc = '\x1b';
if (strchr(path, esc))
{
apr_size_t len = strlen(path);
apr_size_t i;
svn_stringbuf_t *unescaped = svn_stringbuf_create_ensure(len,
result_pool);
for (i = 0; i < len; ++i)
if (path[i] == esc)
svn_stringbuf_appendbyte(unescaped, path[++i] + 1 - 'A');
else
svn_stringbuf_appendbyte(unescaped, path[i]);
return unescaped->data;
}
return path;
}
/* Take the ORIGINAL string and replace all occurrences of ":" without
* limiting the key space. Allocate the result in POOL.
*/
static const char *
normalize_key_part(const char *original,
apr_pool_t *pool)
{
apr_size_t i;
apr_size_t len = strlen(original);
svn_stringbuf_t *normalized = svn_stringbuf_create_ensure(len, pool);
for (i = 0; i < len; ++i)
{
char c = original[i];
switch (c)
{
case ':': svn_stringbuf_appendbytes(normalized, "%_", 2);
break;
case '%': svn_stringbuf_appendbytes(normalized, "%%", 2);
break;
default : svn_stringbuf_appendbyte(normalized, c);
}
}
return normalized->data;
}
/* Process the strace output stored in FILE.
*/
static void
parse_file(apr_file_t *file)
{
apr_pool_t *pool = svn_pool_create(NULL);
apr_pool_t *iterpool = svn_pool_create(pool);
/* limit lines to 4k (usually, we need less than 200 bytes) */
svn_stringbuf_t *line = svn_stringbuf_create_ensure(4096, pool);
do
{
svn_error_t *err = NULL;
line->len = line->blocksize-1;
err = svn_io_read_length_line(file, line->data, &line->len, iterpool);
svn_error_clear(err);
if (err)
break;
parse_line(line);
svn_pool_clear(iterpool);
}
while (line->len > 0);
}
/* A helper function to parse svn:mergeinfo diffs.
*
* These diffs use a special pretty-print format, for instance:
*
* Added: svn:mergeinfo
* ## -0,0 +0,1 ##
* Merged /trunk:r2-3
*
* The hunk header has the following format:
* ## -0,NUMBER_OF_REVERSE_MERGES +0,NUMBER_OF_FORWARD_MERGES ##
*
* At this point, the number of reverse merges has already been
* parsed into HUNK->ORIGINAL_LENGTH, and the number of forward
* merges has been parsed into HUNK->MODIFIED_LENGTH.
*
* The header is followed by a list of mergeinfo, one path per line.
* This function parses such lines. Lines describing reverse merges
* appear first, and then all lines describing forward merges appear.
*
* Parts of the line are affected by i18n. The words 'Merged'
* and 'Reverse-merged' can appear in any language and at any
* position within the line. We can only assume that a leading
* '/' starts the merge source path, the path is followed by
* ":r", which in turn is followed by a mergeinfo revision range,
* which is terminated by whitespace or end-of-string.
*
* If the current line meets the above criteria and we're able
* to parse valid mergeinfo from it, the resulting mergeinfo
* is added to patch->mergeinfo or patch->reverse_mergeinfo,
* and we proceed to the next line.
*/
static svn_error_t *
parse_mergeinfo(svn_boolean_t *found_mergeinfo,
svn_stringbuf_t *line,
svn_diff_hunk_t *hunk,
svn_patch_t *patch,
apr_pool_t *result_pool,
apr_pool_t *scratch_pool)
{
char *slash = strchr(line->data, '/');
char *colon = strrchr(line->data, ':');
*found_mergeinfo = FALSE;
if (slash && colon && colon[1] == 'r' && slash < colon)
{
svn_stringbuf_t *input;
svn_mergeinfo_t mergeinfo = NULL;
char *s;
svn_error_t *err;
input = svn_stringbuf_create_ensure(line->len, scratch_pool);
/* Copy the merge source path + colon */
s = slash;
while (s <= colon)
{
svn_stringbuf_appendbyte(input, *s);
s++;
}
/* skip 'r' after colon */
s++;
/* Copy the revision range. */
while (s < line->data + line->len)
{
if (svn_ctype_isspace(*s))
break;
svn_stringbuf_appendbyte(input, *s);
s++;
}
err = svn_mergeinfo_parse(&mergeinfo, input->data, result_pool);
if (err && err->apr_err == SVN_ERR_MERGEINFO_PARSE_ERROR)
{
svn_error_clear(err);
mergeinfo = NULL;
}
else
SVN_ERR(err);
if (mergeinfo)
{
if (hunk->original_length > 0) /* reverse merges */
{
if (patch->reverse)
{
if (patch->mergeinfo == NULL)
patch->mergeinfo = mergeinfo;
else
SVN_ERR(svn_mergeinfo_merge2(patch->mergeinfo,
mergeinfo,
result_pool,
scratch_pool));
}
else
{
if (patch->reverse_mergeinfo == NULL)
patch->reverse_mergeinfo = mergeinfo;
else
SVN_ERR(svn_mergeinfo_merge2(patch->reverse_mergeinfo,
mergeinfo,
result_pool,
scratch_pool));
}
hunk->original_length--;
}
else if (hunk->modified_length > 0) /* forward merges */
{
if (patch->reverse)
{
if (patch->reverse_mergeinfo == NULL)
patch->reverse_mergeinfo = mergeinfo;
else
SVN_ERR(svn_mergeinfo_merge2(patch->reverse_mergeinfo,
mergeinfo,
result_pool,
scratch_pool));
}
else
{
if (patch->mergeinfo == NULL)
patch->mergeinfo = mergeinfo;
else
SVN_ERR(svn_mergeinfo_merge2(patch->mergeinfo,
mergeinfo,
result_pool,
scratch_pool));
}
hunk->modified_length--;
}
*found_mergeinfo = TRUE;
}
}
return SVN_NO_ERROR;
}
/* Try to parse a hunk header in string HEADER, putting parsed information
* into HUNK. Return TRUE if the header parsed correctly. ATAT is the
* character string used to delimit the hunk header.
* Do all allocations in POOL. */
static svn_boolean_t
parse_hunk_header(const char *header, svn_diff_hunk_t *hunk,
const char *atat, apr_pool_t *pool)
{
const char *p;
const char *start;
svn_stringbuf_t *range;
p = header + strlen(atat);
if (*p != ' ')
/* No. */
return FALSE;
p++;
if (*p != '-')
/* Nah... */
return FALSE;
/* OK, this may be worth allocating some memory for... */
range = svn_stringbuf_create_ensure(31, pool);
start = ++p;
while (*p && *p != ' ')
{
p++;
}
if (*p != ' ')
/* No no no... */
return FALSE;
svn_stringbuf_appendbytes(range, start, p - start);
/* Try to parse the first range. */
if (! parse_range(&hunk->original_start, &hunk->original_length, range->data))
return FALSE;
/* Clear the stringbuf so we can reuse it for the second range. */
svn_stringbuf_setempty(range);
p++;
if (*p != '+')
/* Eeek! */
return FALSE;
/* OK, this may be worth copying... */
start = ++p;
while (*p && *p != ' ')
{
p++;
}
if (*p != ' ')
/* No no no... */
return FALSE;
svn_stringbuf_appendbytes(range, start, p - start);
/* Check for trailing @@ */
p++;
if (! starts_with(p, atat))
return FALSE;
/* There may be stuff like C-function names after the trailing @@,
* but we ignore that. */
/* Try to parse the second range. */
if (! parse_range(&hunk->modified_start, &hunk->modified_length, range->data))
return FALSE;
/* Hunk header is good. */
return TRUE;
}
svn_stringbuf_t *
svn_stringbuf_create_empty(apr_pool_t *pool)
{
return svn_stringbuf_create_ensure(0, pool);
}
inline Try<Diff> diff(const std::string& from, const std::string& to)
{
// Initialize the Apache Portable Runtime subsystem, as necessary
// for using the svn library.
initialize();
// Note that svn_pool_create wraps apr_pool_create_ex, which is
// thread safe, see: http://goo.gl/NX0hps.
apr_pool_t* pool = svn_pool_create(NULL);
// First we need to produce a text delta stream by diffing 'source'
// against 'target'.
svn_string_t source;
source.data = from.data();
source.len = from.length();
svn_string_t target;
target.data = to.data();
target.len = to.length();
svn_txdelta_stream_t* delta;
#if SVN_VER_MAJOR >= 1 && SVN_VER_MINOR >= 8
svn_txdelta2(
&delta,
svn_stream_from_string(&source, pool),
svn_stream_from_string(&target, pool),
false,
pool);
#else
svn_txdelta(
&delta,
svn_stream_from_string(&source, pool),
svn_stream_from_string(&target, pool),
pool);
#endif
// Now we want to convert this text delta stream into an svndiff
// format based diff. Setup the handler that will consume the text
// delta and produce the svndiff.
svn_txdelta_window_handler_t handler;
void* baton = NULL;
svn_stringbuf_t* diff = svn_stringbuf_create_ensure(1024, pool);
#if SVN_VER_MAJOR >= 1 && SVN_VER_MINOR >= 7
svn_txdelta_to_svndiff3(
&handler,
&baton,
svn_stream_from_stringbuf(diff, pool),
0,
SVN_DELTA_COMPRESSION_LEVEL_DEFAULT,
pool);
#elif SVN_VER_MAJOR >= 1 && SVN_VER_MINOR >= 4
svn_txdelta_to_svndiff2(
&handler,
&baton,
svn_stream_from_stringbuf(diff, pool),
0,
pool);
#else
svn_txdelta_to_svndiff(
svn_stream_from_stringbuf(diff, pool),
pool,
&handler,
&baton);
#endif
// Now feed the text delta to the handler.
svn_error_t* error = svn_txdelta_send_txstream(delta, handler, baton, pool);
if (error != NULL) {
char buffer[1024];
std::string message(svn_err_best_message(error, buffer, 1024));
svn_pool_destroy(pool);
return Error(message);
}
Diff d(std::string(diff->data, diff->len));
svn_pool_destroy(pool);
return d;
}
/* A helper for reading a line of text from a range in the patch file.
*
* Allocate *STRINGBUF in RESULT_POOL, and read into it one line from FILE.
* Reading stops either after a line-terminator was found or after MAX_LEN
* bytes have been read. The line-terminator is not stored in *STRINGBUF.
*
* The line-terminator is detected automatically and stored in *EOL
* if EOL is not NULL. If EOF is reached and FILE does not end
* with a newline character, and EOL is not NULL, *EOL is set to NULL.
*
* SCRATCH_POOL is used for temporary allocations.
*/
static svn_error_t *
readline(apr_file_t *file,
svn_stringbuf_t **stringbuf,
const char **eol,
svn_boolean_t *eof,
apr_size_t max_len,
apr_pool_t *result_pool,
apr_pool_t *scratch_pool)
{
svn_stringbuf_t *str;
const char *eol_str;
apr_size_t numbytes;
char c;
apr_size_t len;
svn_boolean_t found_eof;
str = svn_stringbuf_create_ensure(80, result_pool);
/* Read bytes into STR up to and including, but not storing,
* the next EOL sequence. */
eol_str = NULL;
numbytes = 1;
len = 0;
found_eof = FALSE;
while (!found_eof)
{
if (len < max_len)
SVN_ERR(svn_io_file_read_full2(file, &c, sizeof(c), &numbytes,
&found_eof, scratch_pool));
len++;
if (numbytes != 1 || len > max_len)
{
found_eof = TRUE;
break;
}
if (c == '\n')
{
eol_str = "\n";
}
else if (c == '\r')
{
eol_str = "\r";
if (!found_eof && len < max_len)
{
apr_off_t pos;
/* Check for "\r\n" by peeking at the next byte. */
pos = 0;
SVN_ERR(svn_io_file_seek(file, APR_CUR, &pos, scratch_pool));
SVN_ERR(svn_io_file_read_full2(file, &c, sizeof(c), &numbytes,
&found_eof, scratch_pool));
if (numbytes == 1 && c == '\n')
{
eol_str = "\r\n";
len++;
}
else
{
/* Pretend we never peeked. */
SVN_ERR(svn_io_file_seek(file, APR_SET, &pos, scratch_pool));
found_eof = FALSE;
numbytes = 1;
}
}
}
else
svn_stringbuf_appendbyte(str, c);
if (eol_str)
break;
}
if (eol)
*eol = eol_str;
if (eof)
*eof = found_eof;
*stringbuf = str;
return SVN_NO_ERROR;
}
static svn_error_t *
stream_readline_chunky(svn_stringbuf_t **stringbuf,
svn_boolean_t *eof,
const char *eol,
svn_stream_t *stream,
apr_pool_t *pool)
{
/* Read larger chunks of data at once into this buffer and scan
* that for EOL. A good chunk size should be about 80 chars since
* most text lines will be shorter. However, don't use a much
* larger value because filling the buffer from the stream takes
* time as well.
*/
char buffer[LINE_CHUNK_SIZE+1];
/* variables */
svn_stream_mark_t *mark;
apr_size_t numbytes;
const char *eol_pos;
apr_size_t total_parsed = 0;
/* invariant for this call */
const size_t eol_len = strlen(eol);
/* Remember the line start so this plus the line length will be
* the position to move to at the end of this function.
*/
SVN_ERR(svn_stream_mark(stream, &mark, pool));
/* Read the first chunk. */
numbytes = LINE_CHUNK_SIZE;
SVN_ERR(svn_stream_read(stream, buffer, &numbytes));
buffer[numbytes] = '\0';
/* Look for the EOL in this first chunk. If we find it, we are done here.
*/
eol_pos = strstr(buffer, eol);
if (eol_pos != NULL)
{
*stringbuf = svn_stringbuf_ncreate(buffer, eol_pos - buffer, pool);
total_parsed = eol_pos - buffer + eol_len;
}
else if (numbytes < LINE_CHUNK_SIZE)
{
/* We hit EOF but not EOL.
*/
*stringbuf = svn_stringbuf_ncreate(buffer, numbytes, pool);
*eof = TRUE;
return SVN_NO_ERROR;
}
else
{
/* A larger buffer for the string is needed. */
svn_stringbuf_t *str;
str = svn_stringbuf_create_ensure(2*LINE_CHUNK_SIZE, pool);
svn_stringbuf_appendbytes(str, buffer, numbytes);
*stringbuf = str;
/* Loop reading chunks until an EOL was found. If we hit EOF, fall
* back to the standard implementation. */
do
{
/* Append the next chunk to the string read so far.
*/
svn_stringbuf_ensure(str, str->len + LINE_CHUNK_SIZE);
numbytes = LINE_CHUNK_SIZE;
SVN_ERR(svn_stream_read(stream, str->data + str->len, &numbytes));
str->len += numbytes;
str->data[str->len] = '\0';
/* Look for the EOL in the new data plus the last part of the
* previous chunk because the EOL may span over the boundary
* between both chunks.
*/
eol_pos = strstr(str->data + str->len - numbytes - (eol_len-1), eol);
if ((numbytes < LINE_CHUNK_SIZE) && (eol_pos == NULL))
{
/* We hit EOF instead of EOL. */
*eof = TRUE;
return SVN_NO_ERROR;
}
}
while (eol_pos == NULL);
/* Number of bytes we actually consumed (i.e. line + EOF).
* We need to "return" the rest to the stream by moving its
* read pointer.
*/
total_parsed = eol_pos - str->data + eol_len;
/* Terminate the string at the EOL postion and return it. */
str->len = eol_pos - str->data;
str->data[str->len] = 0;
}
/* Move the stream read pointer to the first position behind the EOL.
*/
SVN_ERR(svn_stream_seek(stream, mark));
return svn_error_trace(svn_stream_skip(stream, total_parsed));
}
/* This helper is the main "meat" of the editor -- it does all the
work of writing a node record.
Write out a node record for PATH of type KIND under EB->FS_ROOT.
ACTION describes what is happening to the node (see enum svn_node_action).
Write record to writable EB->STREAM, using EB->BUFFER to write in chunks.
If the node was itself copied, IS_COPY is TRUE and the
path/revision of the copy source are in CMP_PATH/CMP_REV. If
IS_COPY is FALSE, yet CMP_PATH/CMP_REV are valid, this node is part
of a copied subtree.
*/
static svn_error_t *
dump_node(struct edit_baton *eb,
const char *path,
svn_node_kind_t kind,
enum svn_node_action action,
svn_boolean_t is_copy,
const char *cmp_path,
svn_revnum_t cmp_rev,
apr_pool_t *pool)
{
svn_stringbuf_t *propstring;
svn_filesize_t content_length = 0;
apr_size_t len;
svn_boolean_t must_dump_text = FALSE, must_dump_props = FALSE;
const char *compare_path = path;
svn_revnum_t compare_rev = eb->current_rev - 1;
svn_fs_root_t *compare_root = NULL;
apr_file_t *delta_file = NULL;
/* Maybe validate the path. */
if (eb->verify || eb->notify_func)
{
svn_error_t *err = svn_fs__path_valid(path, pool);
if (err)
{
if (eb->notify_func)
{
char errbuf[512]; /* ### svn_strerror() magic number */
svn_repos_notify_t *notify;
notify = svn_repos_notify_create(svn_repos_notify_warning, pool);
notify->warning = svn_repos_notify_warning_invalid_fspath;
notify->warning_str = apr_psprintf(
pool,
_("E%06d: While validating fspath '%s': %s"),
err->apr_err, path,
svn_err_best_message(err, errbuf, sizeof(errbuf)));
eb->notify_func(eb->notify_baton, notify, pool);
}
/* Return the error in addition to notifying about it. */
if (eb->verify)
return svn_error_trace(err);
else
svn_error_clear(err);
}
}
/* Write out metadata headers for this file node. */
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_PATH ": %s\n",
path));
if (kind == svn_node_file)
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_KIND ": file\n"));
else if (kind == svn_node_dir)
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_KIND ": dir\n"));
/* Remove leading slashes from copyfrom paths. */
if (cmp_path)
cmp_path = svn_relpath_canonicalize(cmp_path, pool);
/* Validate the comparison path/rev. */
if (ARE_VALID_COPY_ARGS(cmp_path, cmp_rev))
{
compare_path = cmp_path;
compare_rev = cmp_rev;
}
if (action == svn_node_action_change)
{
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_ACTION
": change\n"));
/* either the text or props changed, or possibly both. */
SVN_ERR(svn_fs_revision_root(&compare_root,
svn_fs_root_fs(eb->fs_root),
compare_rev, pool));
SVN_ERR(svn_fs_props_changed(&must_dump_props,
compare_root, compare_path,
eb->fs_root, path, pool));
if (kind == svn_node_file)
SVN_ERR(svn_fs_contents_changed(&must_dump_text,
compare_root, compare_path,
eb->fs_root, path, pool));
}
else if (action == svn_node_action_replace)
{
if (! is_copy)
{
/* a simple delete+add, implied by a single 'replace' action. */
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_ACTION
": replace\n"));
/* definitely need to dump all content for a replace. */
if (kind == svn_node_file)
must_dump_text = TRUE;
must_dump_props = TRUE;
}
else
{
/* more complex: delete original, then add-with-history. */
/* the path & kind headers have already been printed; just
add a delete action, and end the current record.*/
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_ACTION
": delete\n\n"));
/* recurse: print an additional add-with-history record. */
SVN_ERR(dump_node(eb, path, kind, svn_node_action_add,
is_copy, compare_path, compare_rev, pool));
/* we can leave this routine quietly now, don't need to dump
any content; that was already done in the second record. */
must_dump_text = FALSE;
must_dump_props = FALSE;
}
}
else if (action == svn_node_action_delete)
{
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_ACTION
": delete\n"));
/* we can leave this routine quietly now, don't need to dump
any content. */
must_dump_text = FALSE;
must_dump_props = FALSE;
}
else if (action == svn_node_action_add)
{
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_ACTION ": add\n"));
if (! is_copy)
{
/* Dump all contents for a simple 'add'. */
if (kind == svn_node_file)
must_dump_text = TRUE;
must_dump_props = TRUE;
}
else
{
if (!eb->verify && cmp_rev < eb->oldest_dumped_rev
&& eb->notify_func)
{
svn_repos_notify_t *notify =
svn_repos_notify_create(svn_repos_notify_warning, pool);
notify->warning = svn_repos_notify_warning_found_old_reference;
notify->warning_str = apr_psprintf(
pool,
_("Referencing data in revision %ld,"
" which is older than the oldest"
" dumped revision (r%ld). Loading this dump"
" into an empty repository"
" will fail."),
cmp_rev, eb->oldest_dumped_rev);
eb->found_old_reference = TRUE;
eb->notify_func(eb->notify_baton, notify, pool);
}
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_NODE_COPYFROM_REV
": %ld\n"
SVN_REPOS_DUMPFILE_NODE_COPYFROM_PATH
": %s\n",
cmp_rev, cmp_path));
SVN_ERR(svn_fs_revision_root(&compare_root,
svn_fs_root_fs(eb->fs_root),
compare_rev, pool));
/* Need to decide if the copied node had any extra textual or
property mods as well. */
SVN_ERR(svn_fs_props_changed(&must_dump_props,
compare_root, compare_path,
eb->fs_root, path, pool));
if (kind == svn_node_file)
{
svn_checksum_t *checksum;
const char *hex_digest;
SVN_ERR(svn_fs_contents_changed(&must_dump_text,
compare_root, compare_path,
eb->fs_root, path, pool));
SVN_ERR(svn_fs_file_checksum(&checksum, svn_checksum_md5,
compare_root, compare_path,
FALSE, pool));
hex_digest = svn_checksum_to_cstring(checksum, pool);
if (hex_digest)
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_TEXT_COPY_SOURCE_MD5
": %s\n", hex_digest));
SVN_ERR(svn_fs_file_checksum(&checksum, svn_checksum_sha1,
compare_root, compare_path,
FALSE, pool));
hex_digest = svn_checksum_to_cstring(checksum, pool);
if (hex_digest)
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_TEXT_COPY_SOURCE_SHA1
": %s\n", hex_digest));
}
}
}
if ((! must_dump_text) && (! must_dump_props))
{
/* If we're not supposed to dump text or props, so be it, we can
just go home. However, if either one needs to be dumped,
then our dumpstream format demands that at a *minimum*, we
see a lone "PROPS-END" as a divider between text and props
content within the content-block. */
len = 2;
return svn_stream_write(eb->stream, "\n\n", &len); /* ### needed? */
}
/*** Start prepping content to dump... ***/
/* If we are supposed to dump properties, write out a property
length header and generate a stringbuf that contains those
property values here. */
if (must_dump_props)
{
apr_hash_t *prophash, *oldhash = NULL;
apr_size_t proplen;
svn_stream_t *propstream;
SVN_ERR(svn_fs_node_proplist(&prophash, eb->fs_root, path, pool));
/* If this is a partial dump, then issue a warning if we dump mergeinfo
properties that refer to revisions older than the first revision
dumped. */
if (!eb->verify && eb->notify_func && eb->oldest_dumped_rev > 1)
{
svn_string_t *mergeinfo_str = apr_hash_get(prophash,
SVN_PROP_MERGEINFO,
APR_HASH_KEY_STRING);
if (mergeinfo_str)
{
svn_mergeinfo_t mergeinfo, old_mergeinfo;
SVN_ERR(svn_mergeinfo_parse(&mergeinfo, mergeinfo_str->data,
pool));
SVN_ERR(svn_mergeinfo__filter_mergeinfo_by_ranges(
&old_mergeinfo, mergeinfo,
eb->oldest_dumped_rev - 1, 0,
TRUE, pool, pool));
if (apr_hash_count(old_mergeinfo))
{
svn_repos_notify_t *notify =
svn_repos_notify_create(svn_repos_notify_warning, pool);
notify->warning = svn_repos_notify_warning_found_old_mergeinfo;
notify->warning_str = apr_psprintf(
pool,
_("Mergeinfo referencing revision(s) prior "
"to the oldest dumped revision (r%ld). "
"Loading this dump may result in invalid "
"mergeinfo."),
eb->oldest_dumped_rev);
eb->found_old_mergeinfo = TRUE;
eb->notify_func(eb->notify_baton, notify, pool);
}
}
}
if (eb->use_deltas && compare_root)
{
/* Fetch the old property hash to diff against and output a header
saying that our property contents are a delta. */
SVN_ERR(svn_fs_node_proplist(&oldhash, compare_root, compare_path,
pool));
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_PROP_DELTA
": true\n"));
}
else
oldhash = apr_hash_make(pool);
propstring = svn_stringbuf_create_ensure(0, pool);
propstream = svn_stream_from_stringbuf(propstring, pool);
SVN_ERR(svn_hash_write_incremental(prophash, oldhash, propstream,
"PROPS-END", pool));
SVN_ERR(svn_stream_close(propstream));
proplen = propstring->len;
content_length += proplen;
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_PROP_CONTENT_LENGTH
": %" APR_SIZE_T_FMT "\n", proplen));
}
/* If we are supposed to dump text, write out a text length header
here, and an MD5 checksum (if available). */
if (must_dump_text && (kind == svn_node_file))
{
svn_checksum_t *checksum;
const char *hex_digest;
svn_filesize_t textlen;
if (eb->use_deltas)
{
/* Compute the text delta now and write it into a temporary
file, so that we can find its length. Output a header
saying our text contents are a delta. */
SVN_ERR(store_delta(&delta_file, &textlen, compare_root,
compare_path, eb->fs_root, path, pool));
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_TEXT_DELTA
": true\n"));
if (compare_root)
{
SVN_ERR(svn_fs_file_checksum(&checksum, svn_checksum_md5,
compare_root, compare_path,
FALSE, pool));
hex_digest = svn_checksum_to_cstring(checksum, pool);
if (hex_digest)
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_TEXT_DELTA_BASE_MD5
": %s\n", hex_digest));
SVN_ERR(svn_fs_file_checksum(&checksum, svn_checksum_sha1,
compare_root, compare_path,
FALSE, pool));
hex_digest = svn_checksum_to_cstring(checksum, pool);
if (hex_digest)
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_TEXT_DELTA_BASE_SHA1
": %s\n", hex_digest));
}
}
else
{
/* Just fetch the length of the file. */
SVN_ERR(svn_fs_file_length(&textlen, eb->fs_root, path, pool));
}
content_length += textlen;
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_TEXT_CONTENT_LENGTH
": %" SVN_FILESIZE_T_FMT "\n", textlen));
SVN_ERR(svn_fs_file_checksum(&checksum, svn_checksum_md5,
eb->fs_root, path, FALSE, pool));
hex_digest = svn_checksum_to_cstring(checksum, pool);
if (hex_digest)
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_TEXT_CONTENT_MD5
": %s\n", hex_digest));
SVN_ERR(svn_fs_file_checksum(&checksum, svn_checksum_sha1,
eb->fs_root, path, FALSE, pool));
hex_digest = svn_checksum_to_cstring(checksum, pool);
if (hex_digest)
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_TEXT_CONTENT_SHA1
": %s\n", hex_digest));
}
/* 'Content-length:' is the last header before we dump the content,
and is the sum of the text and prop contents lengths. We write
this only for the benefit of non-Subversion RFC-822 parsers. */
SVN_ERR(svn_stream_printf(eb->stream, pool,
SVN_REPOS_DUMPFILE_CONTENT_LENGTH
": %" SVN_FILESIZE_T_FMT "\n\n",
content_length));
/* Dump property content if we're supposed to do so. */
if (must_dump_props)
{
len = propstring->len;
SVN_ERR(svn_stream_write(eb->stream, propstring->data, &len));
}
/* Dump text content */
if (must_dump_text && (kind == svn_node_file))
{
svn_stream_t *contents;
if (delta_file)
{
/* Make sure to close the underlying file when the stream is
closed. */
contents = svn_stream_from_aprfile2(delta_file, FALSE, pool);
}
else
SVN_ERR(svn_fs_file_contents(&contents, eb->fs_root, path, pool));
SVN_ERR(svn_stream_copy3(contents, svn_stream_disown(eb->stream, pool),
NULL, NULL, pool));
}
len = 2;
return svn_stream_write(eb->stream, "\n\n", &len); /* ### needed? */
}