convert: add round trip check based on 'core.checkRoundtripEncoding'
UTF supports lossless conversion round tripping and conversions between UTF and other encodings are mostly round trip safe as Unicode aims to be a superset of all other character encodings. However, certain encodings (e.g. SHIFT-JIS) are known to have round trip issues [1]. Add 'core.checkRoundtripEncoding', which contains a comma separated list of encodings, to define for what encodings Git should check the conversion round trip if they are used in the 'working-tree-encoding' attribute. Set SHIFT-JIS as default value for 'core.checkRoundtripEncoding'. [1] https://support.microsoft.com/en-us/help/170559/prb-conversion-problem-between-shift-jis-and-unicode Signed-off-by: Lars Schneider <larsxschneider@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Lars Schneider committed
Apr 15, 2018 at 20:16 UTC
e92d6225361eba5ff34696122d1491dc7ace2a5a
7 files changed
+137
Documentation/config.txt
+6
@@ -528,6 +528,12 @@ core.autocrlf::
528
This variable can be set to 'input',
529
in which case no output conversion is performed.
530
531
+core.checkRoundtripEncoding::
532
+ A comma and/or whitespace separated list of encodings that Git
533
+ performs UTF-8 round trip checks on if they are used in an
534
+ `working-tree-encoding` attribute (see linkgit:gitattributes[5]).
535
+ The default value is `SHIFT-JIS`.
536
+
537
core.symlinks::
538
If false, symbolic links are checked out as small plain files that
539
contain the link text. linkgit:git-update-index[1] and
Documentation/gitattributes.txt
+8
@@ -312,6 +312,14 @@ number of pitfalls:
312
internal contents as UTF-8 and try to convert it to UTF-16 on checkout.
313
That operation will fail and cause an error.
314
315
+- Reencoding content to non-UTF encodings can cause errors as the
316
+ conversion might not be UTF-8 round trip safe. If you suspect your
317
+ encoding to not be round trip safe, then add it to
318
+ `core.checkRoundtripEncoding` to make Git check the round trip
319
+ encoding (see linkgit:git-config[1]). SHIFT-JIS (Japanese character
320
+ set) is known to have round trip issues with UTF-8 and is checked by
321
+ default.
322
+
323
- Reencoding content requires resources that might slow down certain
324
Git operations (e.g 'git checkout' or 'git add').
325
config.c
+5
@@ -1172,6 +1172,11 @@ static int git_default_core_config(const char *var, const char *value)
1172
return 0;
1173
}
1174
1175
+ if (!strcmp(var, "core.checkroundtripencoding")) {
1176
+ check_roundtrip_encoding = xstrdup(value);
1177
+ return 0;
1178
+ }
1179
+
1180
if (!strcmp(var, "core.notesref")) {
1181
notes_ref_name = xstrdup(value);
1182
return 0;
convert.c
+77
@@ -347,6 +347,42 @@ static void trace_encoding(const char *context, const char *path,
347
strbuf_release(&trace);
348
}
349
350
+static int check_roundtrip(const char *enc_name)
351
+{
352
+ /*
353
+ * check_roundtrip_encoding contains a string of comma and/or
354
+ * space separated encodings (eg. "UTF-16, ASCII, CP1125").
355
+ * Search for the given encoding in that string.
356
+ */
357
+ const char *found = strcasestr(check_roundtrip_encoding, enc_name);
358
+ const char *next;
359
+ int len;
360
+ if (!found)
361
+ return 0;
362
+ next = found + strlen(enc_name);
363
+ len = strlen(check_roundtrip_encoding);
364
+ return (found && (
365
+ /*
366
+ * check that the found encoding is at the
367
+ * beginning of check_roundtrip_encoding or
368
+ * that it is prefixed with a space or comma
369
+ */
370
+ found == check_roundtrip_encoding || (
371
+ (isspace(found[-1]) || found[-1] == ',')
372
+ )
373
+ ) && (
374
+ /*
375
+ * check that the found encoding is at the
376
+ * end of check_roundtrip_encoding or
377
+ * that it is suffixed with a space or comma
378
+ */
379
+ next == check_roundtrip_encoding + len || (
380
+ next < check_roundtrip_encoding + len &&
381
+ (isspace(next[0]) || next[0] == ',')
382
+ )
383
+ ));
384
+}
385
+
386
static const char *default_encoding = "UTF-8";
387
388
static int encode_to_git(const char *path, const char *src, size_t src_len,
@@ -395,6 +431,47 @@ static int encode_to_git(const char *path, const char *src, size_t src_len,
431
}
432
trace_encoding("destination", path, default_encoding, dst, dst_len);
433
434
+ /*
435
+ * UTF supports lossless conversion round tripping [1] and conversions
436
+ * between UTF and other encodings are mostly round trip safe as
437
+ * Unicode aims to be a superset of all other character encodings.
438
+ * However, certain encodings (e.g. SHIFT-JIS) are known to have round
439
+ * trip issues [2]. Check the round trip conversion for all encodings
440
+ * listed in core.checkRoundtripEncoding.
441
+ *
442
+ * The round trip check is only performed if content is written to Git.
443
+ * This ensures that no information is lost during conversion to/from
444
+ * the internal UTF-8 representation.
445
+ *
446
+ * Please note, the code below is not tested because I was not able to
447
+ * generate a faulty round trip without an iconv error. Iconv errors
448
+ * are already caught above.
449
+ *
450
+ * [1] http://unicode.org/faq/utf_bom.html#gen2
451
+ * [2] https://support.microsoft.com/en-us/help/170559/prb-conversion-problem-between-shift-jis-and-unicode
452
+ */
453
+ if (die_on_error && check_roundtrip(enc)) {
454
+ char *re_src;
455
+ int re_src_len;
456
+
457
+ re_src = reencode_string_len(dst, dst_len,
458
+ enc, default_encoding,
459
+ &re_src_len);
460
+
461
+ trace_printf("Checking roundtrip encoding for %s...\n", enc);
462
+ trace_encoding("reencoded source", path, enc,
463
+ re_src, re_src_len);
464
+
465
+ if (!re_src || src_len != re_src_len ||
466
+ memcmp(src, re_src, src_len)) {
467
+ const char* msg = _("encoding '%s' from %s to %s and "
468
+ "back is not the same");
469
+ die(msg, path, enc, default_encoding);
470
+ }
471
+
472
+ free(re_src);
473
+ }
474
+
475
strbuf_attach(buf, dst, dst_len, dst_len + 1);
476
return 1;
477
}
convert.h
+1
@@ -56,6 +56,7 @@ struct delayed_checkout {
56
};
57
58
extern enum eol core_eol;
59
+extern char *check_roundtrip_encoding;
60
extern const char *get_cached_convert_stats_ascii(const struct index_state *istate,
61
const char *path);
62
extern const char *get_wt_convert_stats_ascii(const char *path);
environment.c
+1
@@ -50,6 +50,7 @@ int check_replace_refs = 1;
50
char *git_replace_ref_base;
51
enum eol core_eol = EOL_UNSET;
52
int global_conv_flags_eol = CONV_EOL_RNDTRP_WARN;
53
+char *check_roundtrip_encoding = "SHIFT-JIS";
54
unsigned whitespace_rule_cfg = WS_DEFAULT_RULE;
55
enum branch_track git_branch_track = BRANCH_TRACK_REMOTE;
56
enum rebase_setup_type autorebase = AUTOREBASE_NEVER;
t/t0028-working-tree-encoding.sh
+39
@@ -203,4 +203,43 @@ test_expect_success 'error if encoding garbage is already in Git' '
203
test_i18ngrep "error: BOM is required" err.out
204
'
205
206
+test_expect_success 'check roundtrip encoding' '
207
+ test_when_finished "rm -f roundtrip.shift roundtrip.utf16" &&
208
+ test_when_finished "git reset --hard HEAD" &&
209
+
210
+ text="hallo there!\nroundtrip test here!" &&
211
+ printf "$text" | iconv -f UTF-8 -t SHIFT-JIS >roundtrip.shift &&
212
+ printf "$text" | iconv -f UTF-8 -t UTF-16 >roundtrip.utf16 &&
213
+ echo "*.shift text working-tree-encoding=SHIFT-JIS" >>.gitattributes &&
214
+
215
+ # SHIFT-JIS encoded files are round-trip checked by default...
216
+ GIT_TRACE=1 git add .gitattributes roundtrip.shift 2>&1 |
217
+ grep "Checking roundtrip encoding for SHIFT-JIS" &&
218
+ git reset &&
219
+
220
+ # ... unless we overwrite the Git config!
221
+ ! GIT_TRACE=1 git -c core.checkRoundtripEncoding=garbage \
222
+ add .gitattributes roundtrip.shift 2>&1 |
223
+ grep "Checking roundtrip encoding for SHIFT-JIS" &&
224
+ git reset &&
225
+
226
+ # UTF-16 encoded files should not be round-trip checked by default...
227
+ ! GIT_TRACE=1 git add roundtrip.utf16 2>&1 |
228
+ grep "Checking roundtrip encoding for UTF-16" &&
229
+ git reset &&
230
+
231
+ # ... unless we tell Git to check it!
232
+ GIT_TRACE=1 git -c core.checkRoundtripEncoding="UTF-16, UTF-32" \
233
+ add roundtrip.utf16 2>&1 |
234
+ grep "Checking roundtrip encoding for utf-16" &&
235
+ git reset &&
236
+
237
+ # ... unless we tell Git to check it!
238
+ # (here we also check that the casing of the encoding is irrelevant)
239
+ GIT_TRACE=1 git -c core.checkRoundtripEncoding="UTF-32, utf-16" \
240
+ add roundtrip.utf16 2>&1 |
241
+ grep "Checking roundtrip encoding for utf-16" &&
242
+ git reset
243
+'
244
+
245
test_done