diff options
Diffstat (limited to 'subversion/libsvn_diff/diff_memory.c')
| -rw-r--r-- | subversion/libsvn_diff/diff_memory.c | 269 |
1 files changed, 156 insertions, 113 deletions
diff --git a/subversion/libsvn_diff/diff_memory.c b/subversion/libsvn_diff/diff_memory.c index 91c197a..00f4c7f 100644 --- a/subversion/libsvn_diff/diff_memory.c +++ b/subversion/libsvn_diff/diff_memory.c @@ -27,6 +27,8 @@ #include <apr_want.h> #include <apr_tables.h> +#include <assert.h> + #include "svn_diff.h" #include "svn_pools.h" #include "svn_types.h" @@ -35,6 +37,7 @@ #include "diff.h" #include "svn_private_config.h" #include "private/svn_adler32.h" +#include "private/svn_diff_private.h" typedef struct source_tokens_t { @@ -123,7 +126,7 @@ datasource_get_next_token(apr_uint32_t *hash, void **token, void *baton, diff_mem_baton_t *mem_baton = baton; source_tokens_t *src = &(mem_baton->sources[datasource_to_index(datasource)]); - if (src->tokens->nelts > src->next_token) + if ((apr_size_t)src->tokens->nelts > src->next_token) { /* There are actually tokens to be returned */ char *buf = mem_baton->normalization_buf[0]; @@ -341,7 +344,8 @@ typedef enum unified_output_e { unified_output_context = 0, unified_output_delete, - unified_output_insert + unified_output_insert, + unified_output_skip } unified_output_e; /* Baton for generating unified diffs */ @@ -350,7 +354,7 @@ typedef struct unified_output_baton_t svn_stream_t *output_stream; const char *header_encoding; source_tokens_t sources[2]; /* 0 == original; 1 == modified */ - apr_off_t next_token; /* next token in original source */ + apr_off_t current_token[2]; /* current token per source */ /* Cached markers, in header_encoding, indexed using unified_output_e */ @@ -363,6 +367,9 @@ typedef struct unified_output_baton_t /* The delimiters of the hunk header, '@@' for text hunks and '##' for * property hunks. */ const char *hunk_delimiter; + /* The string to print after a line that does not end with a newline. + * It must start with a '\'. Typically "\ No newline at end of file". */ + const char *no_newline_string; /* Pool for allocation of temporary memory in the callbacks Should be cleared on entry of each iteration of a callback */ @@ -378,31 +385,28 @@ static svn_error_t * output_unified_token_range(output_baton_t *btn, int tokens, unified_output_e type, - apr_off_t first, - apr_off_t past_last) + apr_off_t until) { source_tokens_t *source = &btn->sources[tokens]; - apr_off_t idx; - - past_last = (past_last > source->tokens->nelts) - ? source->tokens->nelts : past_last; - if (tokens == 0) - /* We get context from the original source, don't expect - to be asked to output a block which starts before - what we already have written. */ - first = (first < btn->next_token) ? btn->next_token : first; + if (until > source->tokens->nelts) + until = source->tokens->nelts; - if (first >= past_last) + if (until <= btn->current_token[tokens]) return SVN_NO_ERROR; /* Do the loop with prefix and token */ - for (idx = first; idx < past_last; idx++) + while (TRUE) { - svn_string_t *token - = APR_ARRAY_IDX(source->tokens, idx, svn_string_t *); - svn_stringbuf_appendcstr(btn->hunk, btn->prefix_str[type]); - svn_stringbuf_appendbytes(btn->hunk, token->data, token->len); + svn_string_t *token = + APR_ARRAY_IDX(source->tokens, btn->current_token[tokens], + svn_string_t *); + + if (type != unified_output_skip) + { + svn_stringbuf_appendcstr(btn->hunk, btn->prefix_str[type]); + svn_stringbuf_appendbytes(btn->hunk, token->data, token->len); + } if (type == unified_output_context) { @@ -411,25 +415,28 @@ output_unified_token_range(output_baton_t *btn, } else if (type == unified_output_delete) btn->hunk_length[0]++; - else + else if (type == unified_output_insert) btn->hunk_length[1]++; + /* ### TODO: Add skip processing for -p handling? */ + + btn->current_token[tokens]++; + if (btn->current_token[tokens] == until) + break; } - if (past_last == source->tokens->nelts && source->ends_without_eol) + + if (btn->current_token[tokens] == source->tokens->nelts + && source->ends_without_eol) { const char *out_str; - SVN_ERR(svn_utf_cstring_from_utf8_ex2 - (&out_str, - /* The string below is intentionally not marked for translation: - it's vital to correct operation of the diff(1)/patch(1) - program pair. */ - APR_EOL_STR "\\ No newline at end of file" APR_EOL_STR, - btn->header_encoding, btn->pool)); + + SVN_ERR(svn_utf_cstring_from_utf8_ex2( + &out_str, btn->no_newline_string, + btn->header_encoding, btn->pool)); svn_stringbuf_appendcstr(btn->hunk, out_str); } - if (tokens == 0) - btn->next_token = past_last; + return SVN_NO_ERROR; } @@ -444,6 +451,8 @@ output_unified_flush_hunk(output_baton_t *baton, { apr_off_t target_token; apr_size_t hunk_len; + apr_off_t old_start; + apr_off_t new_start; if (svn_stringbuf_isempty(baton->hunk)) return SVN_NO_ERROR; @@ -452,58 +461,40 @@ output_unified_flush_hunk(output_baton_t *baton, /* Write the trailing context */ target_token = baton->hunk_start[0] + baton->hunk_length[0] - + SVN_DIFF__UNIFIED_CONTEXT_SIZE; + + SVN_DIFF__UNIFIED_CONTEXT_SIZE; SVN_ERR(output_unified_token_range(baton, 0 /*original*/, unified_output_context, - baton->next_token, target_token)); + target_token)); if (hunk_delimiter == NULL) hunk_delimiter = "@@"; + old_start = baton->hunk_start[0]; + new_start = baton->hunk_start[1]; + + /* If the file is non-empty, convert the line indexes from + zero based to one based */ + if (baton->hunk_length[0]) + old_start++; + if (baton->hunk_length[1]) + new_start++; + /* Write the hunk header */ - if (baton->hunk_length[0] > 0) - /* Convert our 0-based line numbers into unidiff 1-based numbers */ - baton->hunk_start[0]++; - SVN_ERR(svn_stream_printf_from_utf8( - baton->output_stream, baton->header_encoding, - baton->pool, - /* Hunk length 1 is implied, don't show the - length field if we have a hunk that long */ - (baton->hunk_length[0] == 1) - ? ("%s -%" APR_OFF_T_FMT) - : ("%s -%" APR_OFF_T_FMT ",%" APR_OFF_T_FMT), - hunk_delimiter, - baton->hunk_start[0], baton->hunk_length[0])); - - if (baton->hunk_length[1] > 0) - /* Convert our 0-based line numbers into unidiff 1-based numbers */ - baton->hunk_start[1]++; - - - /* Hunk length 1 is implied, don't show the - length field if we have a hunk that long */ - if (baton->hunk_length[1] == 1) - { - SVN_ERR(svn_stream_printf_from_utf8( - baton->output_stream, baton->header_encoding, - baton->pool, - " +%" APR_OFF_T_FMT " %s" APR_EOL_STR, - baton->hunk_start[1], hunk_delimiter)); - } - else - { - SVN_ERR(svn_stream_printf_from_utf8( - baton->output_stream, baton->header_encoding, - baton->pool, - " +%" APR_OFF_T_FMT ",%" APR_OFF_T_FMT " %s" APR_EOL_STR, - baton->hunk_start[1], baton->hunk_length[1], - hunk_delimiter)); - } + SVN_ERR(svn_diff__unified_write_hunk_header( + baton->output_stream, baton->header_encoding, hunk_delimiter, + old_start, baton->hunk_length[0], + new_start, baton->hunk_length[1], + NULL /* hunk_extra_context */, + baton->pool)); hunk_len = baton->hunk->len; SVN_ERR(svn_stream_write(baton->output_stream, baton->hunk->data, &hunk_len)); - baton->hunk_length[0] = baton->hunk_length[1] = 0; + /* Prepare for the next hunk */ + baton->hunk_length[0] = 0; + baton->hunk_length[1] = 0; + baton->hunk_start[0] = 0; + baton->hunk_start[1] = 0; svn_stringbuf_setempty(baton->hunk); return SVN_NO_ERROR; @@ -519,38 +510,91 @@ output_unified_diff_modified(void *baton, apr_off_t latest_start, apr_off_t latest_length) { - output_baton_t *btn = baton; - apr_off_t targ_orig, targ_mod; - - targ_orig = original_start - SVN_DIFF__UNIFIED_CONTEXT_SIZE; - targ_orig = (targ_orig < 0) ? 0 : targ_orig; - targ_mod = modified_start; - - /* If the changed ranges are far enough apart (no overlapping or - * connecting context), flush the current hunk. */ - if (btn->next_token + SVN_DIFF__UNIFIED_CONTEXT_SIZE < targ_orig) - SVN_ERR(output_unified_flush_hunk(btn, btn->hunk_delimiter)); - /* Adjust offset if it's not the first hunk. */ - else if (btn->hunk_length[0] != 0) - targ_orig = btn->next_token; - - if (btn->hunk_length[0] == 0 - && btn->hunk_length[1] == 0) + output_baton_t *output_baton = baton; + apr_off_t context_prefix_length; + apr_off_t prev_context_end; + svn_boolean_t init_hunk = FALSE; + + if (original_start > SVN_DIFF__UNIFIED_CONTEXT_SIZE) + context_prefix_length = SVN_DIFF__UNIFIED_CONTEXT_SIZE; + else + context_prefix_length = original_start; + + /* Calculate where the previous hunk will end if we would write it now + (including the necessary context at the end) */ + if (output_baton->hunk_length[0] > 0 || output_baton->hunk_length[1] > 0) { - btn->hunk_start[0] = targ_orig; - btn->hunk_start[1] = targ_mod + targ_orig - original_start; + prev_context_end = output_baton->hunk_start[0] + + output_baton->hunk_length[0] + + SVN_DIFF__UNIFIED_CONTEXT_SIZE; } + else + { + prev_context_end = -1; - SVN_ERR(output_unified_token_range(btn, 0/*original*/, - unified_output_context, - targ_orig, original_start)); - SVN_ERR(output_unified_token_range(btn, 0/*original*/, + if (output_baton->hunk_start[0] == 0 + && (original_length > 0 || modified_length > 0)) + init_hunk = TRUE; + } + + /* If the changed range is far enough from the previous range, flush the current + hunk. */ + { + apr_off_t new_hunk_start = (original_start - context_prefix_length); + + if (output_baton->current_token[0] < new_hunk_start + && prev_context_end <= new_hunk_start) + { + SVN_ERR(output_unified_flush_hunk(output_baton, + output_baton->hunk_delimiter)); + init_hunk = TRUE; + } + else if (output_baton->hunk_length[0] > 0 + || output_baton->hunk_length[1] > 0) + { + /* We extend the current hunk */ + + /* Original: Output the context preceding the changed range */ + SVN_ERR(output_unified_token_range(output_baton, 0 /* original */, + unified_output_context, + original_start)); + } + } + + /* Original: Skip lines until we are at the beginning of the context we want + to display */ + SVN_ERR(output_unified_token_range(output_baton, 0 /* original */, + unified_output_skip, + original_start - context_prefix_length)); + + if (init_hunk) + { + SVN_ERR_ASSERT(output_baton->hunk_length[0] == 0 + && output_baton->hunk_length[1] == 0); + + output_baton->hunk_start[0] = original_start - context_prefix_length; + output_baton->hunk_start[1] = modified_start - context_prefix_length; + } + + /* Modified: Skip lines until we are at the start of the changed range */ + SVN_ERR(output_unified_token_range(output_baton, 1 /* modified */, + unified_output_skip, + modified_start)); + + /* Original: Output the context preceding the changed range */ + SVN_ERR(output_unified_token_range(output_baton, 0 /* original */, + unified_output_context, + original_start)); + + /* Both: Output the changed range */ + SVN_ERR(output_unified_token_range(output_baton, 0 /* original */, unified_output_delete, - original_start, original_start + original_length)); - return output_unified_token_range(btn, 1/*modified*/, unified_output_insert, - modified_start, - modified_start + modified_length); + SVN_ERR(output_unified_token_range(output_baton, 1 /* modified */, + unified_output_insert, + modified_start + modified_length)); + + return SVN_NO_ERROR; } static const svn_diff_output_fns_t mem_output_unified_vtable = @@ -584,8 +628,12 @@ svn_diff_mem_string_output_unified2(svn_stream_t *output_stream, baton.output_stream = output_stream; baton.pool = svn_pool_create(pool); baton.header_encoding = header_encoding; - baton.hunk = svn_stringbuf_create("", pool); + baton.hunk = svn_stringbuf_create_empty(pool); baton.hunk_delimiter = hunk_delimiter; + baton.no_newline_string + = (hunk_delimiter == NULL || strcmp(hunk_delimiter, "##") != 0) + ? APR_EOL_STR SVN_DIFF__NO_NEWLINE_AT_END_OF_FILE APR_EOL_STR + : APR_EOL_STR SVN_DIFF__NO_NEWLINE_AT_END_OF_PROPERTY APR_EOL_STR; SVN_ERR(svn_utf_cstring_from_utf8_ex2 (&(baton.prefix_str[unified_output_context]), " ", @@ -602,12 +650,9 @@ svn_diff_mem_string_output_unified2(svn_stream_t *output_stream, if (with_diff_header) { - SVN_ERR(svn_stream_printf_from_utf8(output_stream, - header_encoding, pool, - "--- %s" APR_EOL_STR - "+++ %s" APR_EOL_STR, - original_header, - modified_header)); + SVN_ERR(svn_diff__unidiff_write_header( + output_stream, header_encoding, + original_header, modified_header, pool)); } SVN_ERR(svn_diff_output(diff, &baton, @@ -680,7 +725,7 @@ typedef struct merge_output_baton_t /* Tokenized source text */ source_tokens_t sources[3]; - apr_off_t next_token; + apr_off_t next_token[3]; /* Markers for marking conflicted sections */ const char *markers[4]; /* 0 = original, 1 = modified, @@ -710,7 +755,7 @@ flush_context_saver(context_saver_t *cs, int i; for (i = 0; i < SVN_DIFF__UNIFIED_CONTEXT_SIZE; i++) { - int slot = (i + cs->next_slot) % SVN_DIFF__UNIFIED_CONTEXT_SIZE; + apr_size_t slot = (i + cs->next_slot) % SVN_DIFF__UNIFIED_CONTEXT_SIZE; if (cs->data[slot]) { apr_size_t len = cs->len[slot]; @@ -808,15 +853,13 @@ output_merge_token_range(apr_size_t *lines_printed_p, static svn_error_t * output_marker_eol(merge_output_baton_t *btn) { - apr_size_t len = strlen(btn->marker_eol); - return svn_stream_write(btn->output_stream, btn->marker_eol, &len); + return svn_stream_puts(btn->output_stream, btn->marker_eol); } static svn_error_t * output_merge_marker(merge_output_baton_t *btn, int idx) { - apr_size_t len = strlen(btn->markers[idx]); - SVN_ERR(svn_stream_write(btn->output_stream, btn->markers[idx], &len)); + SVN_ERR(svn_stream_puts(btn->output_stream, btn->markers[idx])); return output_marker_eol(btn); } @@ -924,7 +967,7 @@ output_conflict_with_context(void *baton, if (btn->output_stream == btn->context_saver->stream) { if (btn->context_saver->total_written > SVN_DIFF__UNIFIED_CONTEXT_SIZE) - SVN_ERR(svn_stream_printf(btn->real_output_stream, btn->pool, "@@\n")); + SVN_ERR(svn_stream_puts(btn->real_output_stream, "@@\n")); SVN_ERR(flush_context_saver(btn->context_saver, btn->real_output_stream)); } |