[ChangeLog]

2005-11-28  Graydon Hoare  <graydon@redhat.com>

	* translate.cxx (var::assert_hist_compatible): New method.
	(var::hist): New method.
	(c_unparser::load_aggregate): New method.
	(hist_op_downcaster): Remove, it was a mistake.
	(expression_is_hist_op): Likewise.
	(c_tmpcounter::visit_print_format): Implement print(@hist(...)).
	(c_unparser::visit_print_format): Likewise.

	* staptree.h (struct print_format): Add optional hist_op* member.
	* staptree.cxx (traversing_visitor::visit_functioncall): Visit
	hist_op if present in print_format.
	(deep_copy_visitor::visit_print_format): Likewise.

	* parse.cxx (parse_symbol): Special case to consume print(@hist(...)).

	* elaborate.cxx (typeresolution_info::visit_arrayindex): Fix type inference bug.
	(typeresolution_info::visit_foreach_loop): Likewise.

	* testsuite/buildok/print_histograms.stp: New test.

[runtime/ChangeLog]

2005-11-28  Graydon Hoare  <graydon@redhat.com>

	* stat-common.c (_stp_stat_print_histogram): Various formatting
	corrections and aesthetic tweaks.
	(__stp_stat_add): Correction to linear bucket underflow cases.

1 files changed, 13 insertions, 0 deletions

diff --git a/parse.cxx b/parse.cxx
index 628fe420..fe10c75c 100644
--- a/parse.cxx
+++ b/parse.cxx
@@ -1988,6 +1988,19 @@ parser::parse_symbol ()
 		  fmt->args.push_back(e);
 		}
 	    }
+	  else if (name == "print" && 
+		   (peek_kw("@hist_linear") || 
+		    peek_kw("@hist_log")))
+	    {
+	      // We have a special case where we recognize
+	      // print(@hist_foo(bar)) as a magic print-the-histogram
+	      // construct. This is sort of gross but it avoids
+	      // promoting histogram references to typeful
+	      // expressions.
+	      fmt->hist = NULL;
+	      t = parse_hist_op_or_bare_name(fmt->hist, name);
+	      assert(fmt->hist);
+	    }
 	  else
 	    {
 	      // If we are not printing with a format string, we permit