[PATCH 5/6] gas: generalize comment character escaping recognition

[Jan Beulich]

PR gas/4572

Generalize what ab1fadc6b2f0 ("PR22714, Assembler preprocessor loses
track of \@") did to always honor escaped comment chars. Use this then
to support escaped /, %, and * operators on x86, when / is a comment
char (to match the Sun assembler's behavior).

gas/
2020-07-XX  Jan Beulich  <jbeulich@suse.com>

	PR gas/4572
	* app.c (last_char): Drop TC_ARM conditional around it.
	(struct app_save): Drop TC_ARM conditional around last_char.
	(app_push, app_pop): Drop TC_ARM conditional from last_char
	accesses.
	(do_scrub_chars): Likewise. Drop TC_ARM conditional from
	backslash-precedes-comment-character check.

---
While I've tested this against a fair set of targets, I'm not going to
claim I'm sure this may not trigger some issue on some exotic one.

I have to admit that I'm not conviced of this handling model, as I
can't help thinking it may mishandle sequences of many backslashes
followed by a comment char. I've merely chosen this approach as being
less intrusive than any alternative I could think of. Plus quite likely
there are reasons for doing things this way, rather than treating
backslashes as escapes universally (their uses in macros could be a
problem here, for example).

--- a/gas/app.c
+++ b/gas/app.c
@@ -55,9 +55,8 @@ static const char mri_pseudo[] = ".mri 0
 static const char   symver_pseudo[] = ".symver";
 static const char * symver_state;
 #endif
-#ifdef TC_ARM
+
 static char last_char;
-#endif
 
 static char lex[256];
 static const char symbol_chars[] =
@@ -244,9 +243,7 @@ struct app_save
 #if defined TC_ARM && defined OBJ_ELF
   const char * symver_state;
 #endif
-#ifdef TC_ARM
-  char last_char;
-#endif
+  char         last_char;
 };
 
 char *
@@ -276,9 +273,7 @@ app_push (void)
 #if defined TC_ARM && defined OBJ_ELF
   saved->symver_state = symver_state;
 #endif
-#ifdef TC_ARM
   saved->last_char = last_char;
-#endif
 
   /* do_scrub_begin() is not useful, just wastes time.  */
 
@@ -318,9 +313,7 @@ app_pop (char *arg)
 #if defined TC_ARM && defined OBJ_ELF
   symver_state = saved->symver_state;
 #endif
-#ifdef TC_ARM
   last_char = saved->last_char;
-#endif
 
   free (arg);
 }
@@ -1289,13 +1282,11 @@ do_scrub_chars (size_t (*get) (char *, s
 	    goto de_fault;
 #endif
 
-#ifdef TC_ARM
-	  /* For the ARM, care is needed not to damage occurrences of \@
-	     by stripping the @ onwards.  Yuck.  */
+	  /* Care is needed not to damage occurrences of \<comment-char>
+	     by stripping the <comment-char> onwards.  Yuck.  */
 	  if ((to > tostart ? to[-1] : last_char) == '\\')
-	    /* Do not treat the @ as a start-of-comment.  */
+	    /* Do not treat the <comment-char> as a start-of-comment.  */
 	    goto de_fault;
-#endif
 
 #ifdef WARN_COMMENTS
 	  if (!found_comment)
@@ -1472,10 +1463,8 @@ do_scrub_chars (size_t (*get) (char *, s
 
  fromeof:
   /* We have reached the end of the input.  */
-#ifdef TC_ARM
   if (to > tostart)
     last_char = to[-1];
-#endif
   return to - tostart;
 
  tofull:
@@ -1489,9 +1478,7 @@ do_scrub_chars (size_t (*get) (char *, s
   else
     saved_input = NULL;
 
-#ifdef TC_ARM
   if (to > tostart)
     last_char = to[-1];
-#endif
   return to - tostart;
 }