ash: fix matching of unicode greek letter rho (cf 81) and similar cases

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

1 files changed, 22 insertions, 1 deletions

diff --git a/shell/ash.c b/shell/ash.c
index 6d46e3719..e5fdd1646 100644
--- a/shell/ash.c
+++ b/shell/ash.c
@@ -5913,6 +5913,7 @@ rmescapes(char *str, int flag)
         while (*p) {
                 if ((unsigned char)*p == CTLQUOTEMARK) {
 // Note: both inquotes and protect_against_glob only affect whether
+// CTLESC,<ch> gets converted to <ch> or to \<ch>
                         inquotes = ~inquotes;
                         p++;
                         protect_against_glob = globbing;
@@ -5925,7 +5926,27 @@ rmescapes(char *str, int flag)
                                 ash_msg_and_raise_error("CTLESC at EOL (shouldn't happen)");
 #endif
                         if (protect_against_glob) {
-                                *q++ = '\\';
+                                /*
+                                 * We used to trust glob() and fnmatch() to eat
+                                 * superfluous escapes (\z where z has no
+                                 * special meaning anyway). But this causes
+                                 * bugs such as string of one greek letter rho
+                                 * (unicode-encoded as two bytes 'cf,81")
+                                 * getting encoded as "cf,CTLESC,81"
+                                 * and here, converted to "cf,\,81" -
+                                 * which does not go well with some flavors
+                                 * of fnmatch() in unicode locales.
+                                 *
+                                 * Lets add "\" only on the chars which need it.
+                                 */
+                                if (*p == '*'
+                                 || *p == '?'
+                                 || *p == '['
+                                /* || *p == ']' maybe also this? */
+                                 || *p == '\\'
+                                ) {
+                                        *q++ = '\\';
+                                }
                         }
                 } else if (*p == '\\' && !inquotes) {
                         /* naked back slash */