From d6248d2640efe3e40a1a9c0bb7bd8903f6beef98 Mon Sep 17 00:00:00 2001 From: John Reiser Date: Fri, 26 Aug 2022 09:51:19 -0700 Subject: [PATCH] archive_entry_pathname() tries UTF-8 if MBS returns EILSEQ For better pathname portability across OS, in particular Windows to Linux. Original bug: unrar in https://bugzilla.redhat.com/show_bug.cgi?id=2120926 modified: libarchive/archive_entry.c --- libarchive/archive_entry.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/libarchive/archive_entry.c b/libarchive/archive_entry.c index ca7a4bdb5..ae6dc3336 100644 --- a/libarchive/archive_entry.c +++ b/libarchive/archive_entry.c @@ -568,6 +568,13 @@ archive_entry_nlink(struct archive_entry *entry) return (entry->ae_stat.aest_nlink); } +/* Instead, our caller could have chosen a specific encoding + * (archive_mstring_get_mbs, archive_mstring_get_utf8, + * archive_mstring_get_wcs). So we should try multiple + * encodings. Try mbs first because of history, even though + * utf8 might be better for pathname portability. + * Also omit wcs because of type mismatch (char * versus wchar *) + */ const char * archive_entry_pathname(struct archive_entry *entry) { @@ -575,6 +582,13 @@ archive_entry_pathname(struct archive_entry *entry) if (archive_mstring_get_mbs( entry->archive, &entry->ae_pathname, &p) == 0) return (p); +#if HAVE_EILSEQ /*{*/ + if (errno == EILSEQ) { + if (archive_mstring_get_utf8( + entry->archive, &entry->ae_pathname, &p) == 0) + return (p); + } +#endif /*}*/ if (errno == ENOMEM) __archive_errx(1, "No memory"); return (NULL);