Re: RFC: fix PR symtab/11464

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Mon, 11 Feb 2013 21:45:01 +0100, Tom Tromey wrote:
> Jan> That's true, that crashes even for me.  I still expected it crashes
> Jan> for you if you remove the two lines:
> Jan>          if (classification == NAME)
> Jan>            break;
> Jan> which cannot / does not crash.
> 
> I thought you were reporting that the test case did not crash before the
> fix.
> 
> Now I don't understand what you are saying here.

It does not matter but the whole time I was trying to say this does not crash
for me:

--- ./gdb/c-exp.y-x	2013-02-11 21:53:30.837153125 +0100
+++ ./gdb/c-exp.y	2013-02-11 21:53:44.947170669 +0100
@@ -2986,8 +2986,8 @@ yylex (void)
 
 	  last_was_coloncolon = 0;
 	  
-	  if (classification == NAME)
-	    break;
+//	  if (classification == NAME)
+//	    break;
 
 	  context_type = yylval.tsym.type;
 	}


> >> I guess you are saying I should invert the if.
> >> Ok, I will do that.
> 
> Jan> No, I did not mean that.  In fact I do not understand much what condition
> Jan> inversion do you mean.
> 
> I mean now the patch reads:
> 
> 	  if (classification != NAME)
> 	    context_type = yylval.tsym.type;
> 
> This still fixes the failing test case, but I haven't run it through a
> complete regression test yet.

In such case there should be:

	  if (classification != NAME)
	    context_type = yylval.tsym.type;
	  else
	    context_type = NULL;

Otherwise context_type could have "stale" content.

But it has no effect on the executed code.


I do not mind the patch form anymore, it was all just about the crashing /
non-crashing misunderstanding.


Thanks,
Jan