This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[3.3 patch] Permission to backport regression fix ?!?


I've asked a couple of times about this.  It's in mainline, and fixes
a bug in cpp -traditional which is present in 3.3 and is a regression from
3.2 (although it's not reported to Bugzilla).

May I please have permission to backport the patch to 3.3 and commit?
The affected code was the same in 3.3 branch and 3.4 mainline, so there 
should be no problem.

	* cpptrad.c (_cpp_scan_out_logical_line): Improve test for
	  whether directive begins at the beginning of a line.
	(testsuite)
	* gcc.dg/cpp/trad/xwin1.c: New test case.

http://gcc.gnu.org/ml/gcc-patches/2003-11/msg01543.html

Here's the patch again:

===================================================================
RCS file: /cvs/gcc/gcc/gcc/cpptrad.c,v
retrieving revision 1.38
retrieving revision 1.39
diff -u -r1.38 -r1.39
--- gcc/gcc/cpptrad.c	2003/11/01 22:56:54	1.38
+++ gcc/gcc/cpptrad.c	2003/11/19 18:48:24	1.39
@@ -350,6 +350,7 @@
   unsigned int c, paren_depth = 0, quote;
   enum ls lex_state = ls_none;
   bool header_ok;
+  const uchar *start_of_input_line;
 
   fmacro.buff = NULL;

@@ -359,6 +360,9 @@
   RLIMIT (pfile->context) = pfile->buffer->rlimit;
   pfile->out.cur = pfile->out.base;
   pfile->out.first_line = pfile->line;
+  /* start_of_input_line is needed to make sure that directives really,
+     really start at the first character of the line. */
+  start_of_input_line = pfile->buffer->cur;
  new_context:
   context = pfile->context;
   cur = CUR (context);
@@ -581,7 +585,7 @@
 	  break;

 	case '#':
-	  if (out - 1 == pfile->out.base
+	  if (cur - 1 == start_of_input_line
 	      /* A '#' from a macro doesn't start a directive.  */
 	      && !pfile->context->prev
 	      && !pfile->state.in_directive)


=============================
And here's the test case again:

/* XWindows (as of 4.3) does some pretty strange things with cpp.
   This tests one of them; the leading comments are supposed to be
   eaten by the preprocessor; but the 'directives' after them are
   supposed to be retained as text, not processed, so that imake's cpp
   can be run on the output!
   { dg-do preprocess }
*/

/**/#if 0
passed
/**/#endif

/* { dg-final { scan-file xwin1.i "(^|\n)#if 0" } } */


-- 
Nathanael Nerode  <neroden at gcc.gnu.org>
http://home.twcny.rr.com/nerode/neroden/fdl.html


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]