submodule: fix case-folding gitdir filesystem collisions
Add a new check when extension.submodulePathConfig is enabled, to detect and prevent case-folding filesystem colisions. When this new check is triggered, a stricter casefolding aware URI encoding is used to percent-encode uppercase characters. By using this check/retry mechanism the uppercase encoding is only applied when necessary, so case-sensitive filesystems are not affected. Signed-off-by: Adrian Ratiu <adrian.ratiu@collabora.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Adrian Ratiu committed
Jan 12, 2026 at 20:46 UTC
1685bba838ace8b4e325616ab914a6b01f18547f
5 files changed
+126
-2
builtin/submodule--helper.c
+25
-1
@@ -473,7 +473,7 @@ static void create_default_gitdir_config(const char *submodule_name)
473
return;
474
}
475
476
- /* Case 2: Try URI-safe (RFC3986) encoding first, this fixes nested gitdirs */
476
+ /* Case 2.1: Try URI-safe (RFC3986) encoding first, this fixes nested gitdirs */
477
strbuf_reset(&gitdir_path);
478
repo_git_path_append(the_repository, &gitdir_path, "modules/");
479
strbuf_addstr_urlencode(&gitdir_path, submodule_name, is_rfc3986_unreserved);
@@ -482,6 +482,30 @@ static void create_default_gitdir_config(const char *submodule_name)
482
return;
483
}
484
485
+ /* Case 2.2: Try extended uppercase URI (RFC3986) encoding, to fix case-folding */
486
+ strbuf_reset(&gitdir_path);
487
+ repo_git_path_append(the_repository, &gitdir_path, "modules/");
488
+ strbuf_addstr_urlencode(&gitdir_path, submodule_name, is_casefolding_rfc3986_unreserved);
489
+ if (!validate_and_set_submodule_gitdir(&gitdir_path, submodule_name))
490
+ return;
491
+
492
+ /* Case 2.3: Try some derived gitdir names, see if one sticks */
493
+ for (char c = '0'; c <= '9'; c++) {
494
+ strbuf_reset(&gitdir_path);
495
+ repo_git_path_append(the_repository, &gitdir_path, "modules/");
496
+ strbuf_addstr_urlencode(&gitdir_path, submodule_name, is_rfc3986_unreserved);
497
+ strbuf_addch(&gitdir_path, c);
498
+ if (!validate_and_set_submodule_gitdir(&gitdir_path, submodule_name))
499
+ return;
500
+
501
+ strbuf_reset(&gitdir_path);
502
+ repo_git_path_append(the_repository, &gitdir_path, "modules/");
503
+ strbuf_addstr_urlencode(&gitdir_path, submodule_name, is_casefolding_rfc3986_unreserved);
504
+ strbuf_addch(&gitdir_path, c);
505
+ if (!validate_and_set_submodule_gitdir(&gitdir_path, submodule_name))
506
+ return;
507
+ }
508
+
509
/* Case 3: nothing worked, error out */
510
die(_("failed to set a valid default config for 'submodule.%s.gitdir'. "
511
"Please ensure it is set, for example by running something like: "
submodule.c
+52
-1
@@ -2254,15 +2254,58 @@ out:
2254
return ret;
2255
}
2256
2257
+static int check_casefolding_conflict(const char *git_dir,
2258
+ const char *submodule_name,
2259
+ const bool suffixes_match)
2260
+{
2261
+ char *p, *modules_dir = xstrdup(git_dir);
2262
+ struct dirent *de;
2263
+ DIR *dir = NULL;
2264
+ int ret = 0;
2265
+
2266
+ if ((p = find_last_dir_sep(modules_dir)))
2267
+ *p = '\0';
2268
+
2269
+ /* No conflict is possible if modules_dir doesn't exist (first clone) */
2270
+ if (!is_directory(modules_dir))
2271
+ goto cleanup;
2272
+
2273
+ dir = opendir(modules_dir);
2274
+ if (!dir) {
2275
+ ret = -1;
2276
+ goto cleanup;
2277
+ }
2278
+
2279
+ /* Check for another directory under .git/modules that differs only in case. */
2280
+ while ((de = readdir(dir))) {
2281
+ if (!strcmp(de->d_name, ".") || !strcmp(de->d_name, ".."))
2282
+ continue;
2283
+
2284
+ if ((suffixes_match || is_git_directory(git_dir)) &&
2285
+ !strcasecmp(de->d_name, submodule_name) &&
2286
+ strcmp(de->d_name, submodule_name)) {
2287
+ ret = -1; /* collision found */
2288
+ break;
2289
+ }
2290
+ }
2291
+
2292
+cleanup:
2293
+ if (dir)
2294
+ closedir(dir);
2295
+ free(modules_dir);
2296
+ return ret;
2297
+}
2298
+
2299
/*
2300
* Encoded gitdir validation, only used when extensions.submodulePathConfig is enabled.
2301
* This does not print errors like the non-encoded version, because encoding is supposed
2302
* to mitigate / fix all these.
2303
*/
2262
-static int validate_submodule_encoded_git_dir(char *git_dir, const char *submodule_name UNUSED)
2304
+static int validate_submodule_encoded_git_dir(char *git_dir, const char *submodule_name)
2305
{
2306
const char *modules_marker = "/modules/";
2307
char *p = git_dir, *last_submodule_name = NULL;
2308
+ int config_ignorecase = 0;
2309
2310
if (!the_repository->repository_format_submodule_path_cfg)
2311
BUG("validate_submodule_encoded_git_dir() must be called with "
@@ -2278,6 +2321,14 @@ static int validate_submodule_encoded_git_dir(char *git_dir, const char *submodu
2321
if (!last_submodule_name || strchr(last_submodule_name, '/'))
2322
return -1;
2323
2324
+ /* Prevent conflicts on case-folding filesystems */
2325
+ repo_config_get_bool(the_repository, "core.ignorecase", &config_ignorecase);
2326
+ if (ignore_case || config_ignorecase) {
2327
+ bool suffixes_match = !strcmp(last_submodule_name, submodule_name);
2328
+ return check_casefolding_conflict(git_dir, submodule_name,
2329
+ suffixes_match);
2330
+ }
2331
+
2332
return 0;
2333
}
2334
t/t7425-submodule-gitdir-path-extension.sh
+35
@@ -403,4 +403,39 @@ test_expect_success 'disabling extensions.submodulePathConfig prevents nested su
403
)
404
'
405
406
+test_expect_success CASE_INSENSITIVE_FS 'verify case-folding conflicts are correctly encoded' '
407
+ git clone -c extensions.submodulePathConfig=true main cloned-folding &&
408
+ (
409
+ cd cloned-folding &&
410
+
411
+ # conflict: the "folding" gitdir will already be taken
412
+ git submodule add ../new-sub "folding" &&
413
+ test_commit lowercase &&
414
+ git submodule add ../new-sub "FoldinG" &&
415
+ test_commit uppercase &&
416
+
417
+ # conflict: the "foo" gitdir will already be taken
418
+ git submodule add ../new-sub "FOO" &&
419
+ test_commit uppercase-foo &&
420
+ git submodule add ../new-sub "foo" &&
421
+ test_commit lowercase-foo &&
422
+
423
+ # create a multi conflict between foobar, fooBar and foo%42ar
424
+ # the "foo" gitdir will already be taken
425
+ git submodule add ../new-sub "foobar" &&
426
+ test_commit lowercase-foobar &&
427
+ git submodule add ../new-sub "foo%42ar" &&
428
+ test_commit encoded-foo%42ar &&
429
+ git submodule add ../new-sub "fooBar" &&
430
+ test_commit mixed-fooBar
431
+ ) &&
432
+ verify_submodule_gitdir_path cloned-folding "folding" "modules/folding" &&
433
+ verify_submodule_gitdir_path cloned-folding "FoldinG" "modules/%46oldin%47" &&
434
+ verify_submodule_gitdir_path cloned-folding "FOO" "modules/FOO" &&
435
+ verify_submodule_gitdir_path cloned-folding "foo" "modules/foo0" &&
436
+ verify_submodule_gitdir_path cloned-folding "foobar" "modules/foobar" &&
437
+ verify_submodule_gitdir_path cloned-folding "foo%42ar" "modules/foo%42ar" &&
438
+ verify_submodule_gitdir_path cloned-folding "fooBar" "modules/fooBar0"
439
+'
440
+
441
test_done
url.c
+7
@@ -9,6 +9,13 @@ int is_rfc3986_unreserved(char ch)
9
ch == '-' || ch == '_' || ch == '.' || ch == '~';
10
}
11
12
+int is_casefolding_rfc3986_unreserved(char c)
13
+{
14
+ return (c >= 'a' && c <= 'z') ||
15
+ (c >= '0' && c <= '9') ||
16
+ c == '-' || c == '.' || c == '_' || c == '~';
17
+}
18
+
19
int is_urlschemechar(int first_flag, int ch)
20
{
21
/*
url.h
+7
@@ -28,4 +28,11 @@ void str_end_url_with_slash(const char *url, char **dest);
28
*/
29
int is_rfc3986_unreserved(char ch);
30
31
+/*
32
+ * This is a variant of is_rfc3986_unreserved() that treats uppercase
33
+ * letters as "reserved". This forces them to be percent-encoded, allowing
34
+ * 'Foo' (%46oo) and 'foo' (foo) to be distinct on case-folding filesystems.
35
+ */
36
+int is_casefolding_rfc3986_unreserved(char c);
37
+
38
#endif /* URL_H */