Handle comments before duplicated msgids in a better way.

5 files changed, 137 insertions, 50 deletions

diff --git a/gettext-tools/src/ChangeLog b/gettext-tools/src/ChangeLog
index f6b3b34..85bafb4 100644
--- a/gettext-tools/src/ChangeLog
+++ b/gettext-tools/src/ChangeLog
@@ -1,3 +1,9 @@
+2003-11-07  Bruno Haible  <bruno@clisp.org>
+
+	* xgettext.c (remember_a_message): Omit the programmer comments of a
+	duplicated msgid only if they are redundant.
+	Reported by Christian Neumair <chris@gnome-de.org>.
+
 2003-11-05  Bruno Haible  <bruno@clisp.org>
 
 	* user-email.sh.in: Renamed from user-email.in. Internationalize, use
diff --git a/gettext-tools/src/xgettext.c b/gettext-tools/src/xgettext.c
index 0e1eabb..328a157 100644
--- a/gettext-tools/src/xgettext.c
+++ b/gettext-tools/src/xgettext.c
@@ -1748,66 +1748,97 @@ meta information, not the empty string.\n")));
      string.  */
   set_format_flags_from_context (is_format, context, mp->msgid, pos, "msgid");
 
-  /* Ask the lexer for the comments it has seen.  Only do this for the
-     first instance, otherwise there could be problems; especially if
-     the same comment appears before each.  */
-  if (!mp->comment_dot)
-    {
-      int j;
-      bool add_all_remaining_comments;
+  /* Ask the lexer for the comments it has seen.  */
+  {
+    size_t nitems_before;
+    size_t nitems_after;
+    int j;
+    bool add_all_remaining_comments;
 
-      add_all_remaining_comments = add_all_comments;
-      for (j = 0; ; ++j)
-	{
-	  const char *s = xgettext_comment (j);
-	  const char *t;
-	  if (s == NULL)
-	    break;
+    nitems_before = (mp->comment_dot != NULL ? mp->comment_dot->nitems : 0);
 
-	  CONVERT_STRING (s);
+    add_all_remaining_comments = add_all_comments;
+    for (j = 0; ; ++j)
+      {
+	const char *s = xgettext_comment (j);
+	const char *t;
+	if (s == NULL)
+	  break;
 
-	  /* To reduce the possibility of unwanted matches we do a two
-	     step match: the line must contain `xgettext:' and one of
-	     the possible format description strings.  */
-	  if ((t = strstr (s, "xgettext:")) != NULL)
-	    {
-	      bool tmp_fuzzy;
-	      enum is_format tmp_format[NFORMATS];
-	      enum is_wrap tmp_wrap;
-	      bool interesting;
+	CONVERT_STRING (s);
 
-	      t += strlen ("xgettext:");
+	/* To reduce the possibility of unwanted matches we do a two
+	   step match: the line must contain `xgettext:' and one of
+	   the possible format description strings.  */
+	if ((t = strstr (s, "xgettext:")) != NULL)
+	  {
+	    bool tmp_fuzzy;
+	    enum is_format tmp_format[NFORMATS];
+	    enum is_wrap tmp_wrap;
+	    bool interesting;
 
-	      po_parse_comment_special (t, &tmp_fuzzy, tmp_format, &tmp_wrap);
+	    t += strlen ("xgettext:");
 
-	      interesting = false;
-	      for (i = 0; i < NFORMATS; i++)
-		if (tmp_format[i] != undecided)
-		  {
-		    is_format[i] = tmp_format[i];
-		    interesting = true;
-		  }
-	      if (tmp_wrap != undecided)
+	    po_parse_comment_special (t, &tmp_fuzzy, tmp_format, &tmp_wrap);
+
+	    interesting = false;
+	    for (i = 0; i < NFORMATS; i++)
+	      if (tmp_format[i] != undecided)
 		{
-		  do_wrap = tmp_wrap;
+		  is_format[i] = tmp_format[i];
 		  interesting = true;
 		}
+	    if (tmp_wrap != undecided)
+	      {
+		do_wrap = tmp_wrap;
+		interesting = true;
+	      }
 
-	      /* If the "xgettext:" marker was followed by an interesting
-		 keyword, and we updated our is_format/do_wrap variables,
-		 we don't print the comment as a #. comment.  */
-	      if (interesting)
-		continue;
-	    }
-	  /* When the comment tag is seen, it drags in not only the line
-	     which it starts, but all remaining comment lines.  */
-	  if (add_all_remaining_comments
-	      || (add_all_remaining_comments =
-		    (comment_tag != NULL
-		     && strncmp (s, comment_tag, strlen (comment_tag)) == 0)))
-	    message_comment_dot_append (mp, s);
-	}
-    }
+	    /* If the "xgettext:" marker was followed by an interesting
+	       keyword, and we updated our is_format/do_wrap variables,
+	       we don't print the comment as a #. comment.  */
+	    if (interesting)
+	      continue;
+	  }
+	/* When the comment tag is seen, it drags in not only the line
+	   which it starts, but all remaining comment lines.  */
+	if (add_all_remaining_comments
+	    || (add_all_remaining_comments =
+		  (comment_tag != NULL
+		   && strncmp (s, comment_tag, strlen (comment_tag)) == 0)))
+	  message_comment_dot_append (mp, s);
+      }
+
+    nitems_after = (mp->comment_dot != NULL ? mp->comment_dot->nitems : 0);
+
+    /* Don't add the comments if they are a repetition of the tail of the
+       already present comments.  This avoids unneeded duplication if the
+       same message appears several times, each time with the same comment.  */
+    if (nitems_before < nitems_after)
+      {
+	size_t added = nitems_after - nitems_before;
+
+	if (added <= nitems_before)
+	  {
+	    bool repeated = true;
+
+	    for (i = 0; i < added; i++)
+	      if (strcmp (mp->comment_dot->item[nitems_before - added + i],
+			  mp->comment_dot->item[nitems_before + i]) != 0)
+		{
+		  repeated = false;
+		  break;
+		}
+
+	    if (repeated)
+	      {
+		for (i = 0; i < added; i++)
+		  free ((char *) mp->comment_dot->item[nitems_before + i]);
+		mp->comment_dot->nitems = nitems_before;
+	      }
+	  }
+      }
+  }
 
   /* If it is not already decided, through programmer comments, whether the
      msgid is a format string, examine the msgid.  This is a heuristic.  */
diff --git a/gettext-tools/tests/ChangeLog b/gettext-tools/tests/ChangeLog
index 140c5d6..6dcee57 100644
--- a/gettext-tools/tests/ChangeLog
+++ b/gettext-tools/tests/ChangeLog
@@ -1,3 +1,8 @@
+2003-11-07  Bruno Haible  <bruno@clisp.org>
+
+	* xgettext-7: New file.
+	* Makefile.am (TESTS): Add it.
+
 2003-11-06  Bruno Haible  <bruno@clisp.org>
 
 	* msgcat-stringtable-1: Fix expected test result.
diff --git a/gettext-tools/tests/Makefile.am b/gettext-tools/tests/Makefile.am
index 8167eb9..bf24bd3 100644
--- a/gettext-tools/tests/Makefile.am
+++ b/gettext-tools/tests/Makefile.am
@@ -56,6 +56,7 @@ TESTS = gettext-1 gettext-2 \
 	msgunfmt-tcl-1 \
 	msguniq-1 msguniq-2 msguniq-3 msguniq-4 \
 	xgettext-1 xgettext-2 xgettext-3 xgettext-4 xgettext-5 xgettext-6 \
+	xgettext-7 \
 	xgettext-c-1 xgettext-c-2 xgettext-c-3 xgettext-c-4 xgettext-c-5 \
 	xgettext-c-6 xgettext-c-7 xgettext-c-8 \
 	xgettext-glade-1 xgettext-glade-2 xgettext-glade-3 \
diff --git a/gettext-tools/tests/xgettext-7 b/gettext-tools/tests/xgettext-7
new file mode 100755
index 0000000..d78b93b
--- /dev/null
+++ b/gettext-tools/tests/xgettext-7
@@ -0,0 +1,44 @@
+#!/bin/sh
+
+# Test of comment extraction in the case of duplicated msgids.
+
+tmpfiles=""
+trap 'rm -fr $tmpfiles' 1 2 3 15
+
+tmpfiles="$tmpfiles xg-test7.c"
+cat <<\EOF > xg-test7.c
+/* first comment */
+/* xgettext: c-format */
+gettext ("abc");
+
+/* first comment */
+/* xgettext: lisp-format */
+gettext ("abc");
+
+/* second comment */
+/* xgettext: python-format */
+gettext ("abc");
+EOF
+
+tmpfiles="$tmpfiles xg-test6.po"
+: ${XGETTEXT=xgettext}
+${XGETTEXT} --omit-header --add-comments -d xg-test7 xg-test7.c
+test $? = 0 || { rm -fr $tmpfiles; exit 1; }
+
+tmpfiles="$tmpfiles xg-test7.ok"
+cat <<\EOF > xg-test7.ok
+#. first comment
+#. second comment
+#: xg-test7.c:3 xg-test7.c:7 xg-test7.c:11
+#, c-format, python-format, lisp-format
+msgid "abc"
+msgstr ""
+EOF
+
+: ${DIFF=diff}
+${DIFF} xg-test7.ok xg-test7.po
+result=$?
+
+rm -fr $tmpfiles
+
+exit $result