Package: release.debian.org Severity: normal Tags: trixie X-Debbugs-Cc: [email protected], Jody Bruchon <[email protected]> Control: affects -1 + src:jdupes User: [email protected] Usertags: pu
[ Reason ] Jody Bruchon, the jdupes author, told me that the hash database feature is disabled due to a conditional logic error introduced in v1.28.0 during the major hashdb changes. The condition was inadvertently written with the opposite logic, completely disabling the hash database functionality. [ Impact ] This does not affect the correctness of jdupes results or pose any data integrity risk, but causes a significant performance regression, especially when processing large, unchanging datasets where the hash database provides substantial performance improvements. Several Debian packages use jdupes during the package build process. [ Tests ] Some manual tests were performed to confirm that the package works fine after fixing the issue. [ Risks ] This is a small and straightforward change provided by upstream, with no expected risks. [ Checklist ] [x] *all* changes are documented in the d/changelog [x] I reviewed all changes and I approve them [x] attach debdiff against the package in (old)stable [x] the issue is verified as fixed in unstable [ Changes ] A small patch provided by upstream was applied. This fix is closing the bug #1148491. [ Other info ] No more info.
diff -Nru jdupes-1.28.0/debian/changelog jdupes-1.28.0/debian/changelog --- jdupes-1.28.0/debian/changelog 2025-10-05 22:45:33.000000000 -0300 +++ jdupes-1.28.0/debian/changelog 2026-09-19 23:18:35.000000000 -0300 @@ -1,3 +1,12 @@ +jdupes (1.28.0-1+deb13u2) trixie; urgency=medium + + * debian/patches/030_fix-hashdb.patch: created to fix a conditional logic + error that disabled the hash database feature, resulting in a significant + performance regression for large, unchanging datasets. Thanks to Jody + Bruchon <[email protected]> (upstream). (Closes: #1148491) + + -- Joao Eriberto Mota Filho <[email protected]> Sat, 19 Sep 2026 23:18:35 -0300 + jdupes (1.28.0-1+deb13u1) trixie; urgency=medium * debian/patches/020_fix-uniq-count.patch: created to fix flag overlap diff -Nru jdupes-1.28.0/debian/patches/030_fix-hashdb.patch jdupes-1.28.0/debian/patches/030_fix-hashdb.patch --- jdupes-1.28.0/debian/patches/030_fix-hashdb.patch 1969-12-31 21:00:00.000000000 -0300 +++ jdupes-1.28.0/debian/patches/030_fix-hashdb.patch 2026-09-19 23:18:35.000000000 -0300 @@ -0,0 +1,63 @@ +Description: Fix a conditional logic error that disables the hash database feature + The incorrect condition was introduced in v1.28.0 and caused the + hash database to be completely disabled. +Author: Jody Bruchon <[email protected]> +Origin: upstream +Bug-Debian: https://bugs.debian.org/1148491 +Last-Update: 2026-09-02 +Index: jdupes-1.28.0/hashdb.c +=================================================================== +--- jdupes-1.28.0.orig/hashdb.c ++++ jdupes-1.28.0/hashdb.c +@@ -530,19 +530,24 @@ int read_hashdb_entry(file_t *file) + if (file == NULL || file->d_name == NULL) goto error_null; + if (get_path_hash(file->d_name, file->d_name_len, &path_hash) != 0) goto error_path_hash; + bucket = path_hash & HT_MASK; +- if (hashdb[bucket] == NULL) return 0; ++ if (hashdb[bucket] == NULL) goto end_read_hashdb; + cur = hashdb[bucket]; + while (1) { + if (cur->path_hash != path_hash) { + if (path_hash < cur->path_hash) cur = cur->left; + else cur = cur->right; +- if (cur == NULL) return 0; ++ if (cur == NULL) goto end_read_hashdb; + continue; + } + /* Found a matching path hash */ +- if (cur->pathlen == file->d_name_len && memcmp(cur->path, file->d_name, cur->pathlen) == 0) { ++ LOUD(fprintf(stderr, "read_hashdb_entry: found a matching path hash\n");) ++ if ((cur->pathlen != file->d_name_len) || (memcmp(cur->path, file->d_name, cur->pathlen) != 0)) { ++ LOUD(fprintf(stderr, "read_hashdb_entry: name check failed, not populating (%u != %u, '%s' != '%s', cmp %d)\n", ++ cur->pathlen, file->d_name_len, ++ cur->path, file->d_name, ++ memcmp(cur->path, file->d_name, cur->pathlen));) + cur = cur->left; +- if (cur == NULL) return 0; ++ if (cur == NULL) goto end_read_hashdb; + continue; + } else { + /* Found a matching path too but check mtime */ +@@ -552,10 +557,12 @@ int read_hashdb_entry(file_t *file) + if (cur->size != file->size) exclude |= 4; + if (exclude != 0) { + /* Invalidate if something has changed */ ++ LOUD(fprintf(stderr, "read_hashdb_entry: metadata changed, invalidating entry\n");) + cur->hashcount = 0; + hashdb_dirty = 1; + return -1; + } ++ LOUD(fprintf(stderr, "read_hashdb_entry: copying hash data to file entry\n");) + file->filehash_partial = cur->partialhash; + if (cur->hashcount == 2) { + file->filehash = cur->fullhash; +@@ -564,6 +571,9 @@ int read_hashdb_entry(file_t *file) + return 1; + } + } ++ ++end_read_hashdb: ++ LOUD(fprintf(stderr, "read_hashdb_entry: exhausted hash scan\n");) + return 0; + + error_null: diff -Nru jdupes-1.28.0/debian/patches/series jdupes-1.28.0/debian/patches/series --- jdupes-1.28.0/debian/patches/series 2025-10-05 22:45:33.000000000 -0300 +++ jdupes-1.28.0/debian/patches/series 2026-09-19 23:18:35.000000000 -0300 @@ -1,2 +1,3 @@ 010_fix-FTBFS-Hurd.patch 020_fix-uniq-count.patch +030_fix-hashdb.patch
