Merge branch 'master' into feature/cmp_scalars

4 files changed, 35 insertions, 9 deletions

diff --git a/support/ChangeLog b/support/ChangeLog
index edef437e..174c65c2 100644
--- a/support/ChangeLog
+++ b/support/ChangeLog
@@ -1,3 +1,7 @@
+2020-11-06         Arnold D. Robbins     <arnold@skeeve.com>
+
+	* dfa.h, regex_internal.c, regex_internal.h: Synced from GNULIB.
+
 2020-11-01         Arnold D. Robbins     <arnold@skeeve.com>
 
 	* dfa.c: Synced from GNULIB.
diff --git a/support/dfa.h b/support/dfa.h
index 2f77f263..65477897 100644
--- a/support/dfa.h
+++ b/support/dfa.h
@@ -18,10 +18,17 @@
 
 /* Written June, 1988 by Mike Haertel */
 
+#ifndef DFA_H_
+#define DFA_H_
+
 #include <regex.h>
 #include <stdbool.h>
 #include <stddef.h>
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
 struct localeinfo; /* See localeinfo.h.  */
 
 /* Element of a list of strings, at least one of which is known to
@@ -131,3 +138,9 @@ extern void dfawarn (const char *);
    takes a single argument, a NUL-terminated string describing the error.
    The user must supply a dfaerror.  */
 extern _Noreturn void dfaerror (const char *);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* dfa.h */
diff --git a/support/regex_internal.c b/support/regex_internal.c
index e1b6b4d5..ed0a1346 100644
--- a/support/regex_internal.c
+++ b/support/regex_internal.c
@@ -300,18 +300,20 @@ build_wcs_upper_buffer (re_string_t *pstr)
       while (byte_idx < end_idx)
 	{
 	  wchar_t wc;
+	  unsigned char ch = pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx];
 
-	  if (isascii (pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx])
-	      && mbsinit (&pstr->cur_state))
+	  if (isascii (ch) && mbsinit (&pstr->cur_state))
 	    {
-	      /* In case of a singlebyte character.  */
-	      pstr->mbs[byte_idx]
-		= toupper (pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx]);
 	      /* The next step uses the assumption that wchar_t is encoded
 		 ASCII-safe: all ASCII values can be converted like this.  */
-	      pstr->wcs[byte_idx] = (wchar_t) pstr->mbs[byte_idx];
-	      ++byte_idx;
-	      continue;
+	      wchar_t wcu = __towupper (ch);
+	      if (isascii (wcu))
+		{
+		  pstr->mbs[byte_idx] = wcu;
+		  pstr->wcs[byte_idx] = wcu;
+		  byte_idx++;
+		  continue;
+		}
 	    }
 
 	  remain_len = end_idx - byte_idx;
@@ -348,7 +350,6 @@ build_wcs_upper_buffer (re_string_t *pstr)
 	    {
 	      /* It is an invalid character, an incomplete character
 		 at the end of the string, or '\0'.  Just use the byte.  */
-	      int ch = pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx];
 	      pstr->mbs[byte_idx] = ch;
 	      /* And also cast it to wide char.  */
 	      pstr->wcs[byte_idx++] = (wchar_t) ch;
diff --git a/support/regex_internal.h b/support/regex_internal.h
index dbc503cd..4a3cf779 100644
--- a/support/regex_internal.h
+++ b/support/regex_internal.h
@@ -77,6 +77,14 @@
 # define isblank(ch) ((ch) == ' ' || (ch) == '\t')
 #endif
 
+/* regex code assumes isascii has its usual numeric meaning,
+   even if the portable character set uses EBCDIC encoding,
+   and even if wint_t is wider than int.  */
+#ifndef _LIBC
+# undef isascii
+# define isascii(c) (((c) & ~0x7f) == 0)
+#endif
+
 #ifdef _LIBC
 # ifndef _RE_DEFINE_LOCALE_FUNCTIONS
 #  define _RE_DEFINE_LOCALE_FUNCTIONS 1