From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 26683 invoked by alias); 11 Feb 2009 21:47:03 -0000 Received: (qmail 26663 invoked by uid 22791); 11 Feb 2009 21:47:01 -0000 X-SWARE-Spam-Status: No, hits=-1.6 required=5.0 tests=AWL,BAYES_00,KAM_STOCKGEN,SPF_HELO_PASS,SPF_PASS X-Spam-Check-By: sourceware.org Received: from mx2.redhat.com (HELO mx2.redhat.com) (66.187.237.31) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Wed, 11 Feb 2009 21:46:54 +0000 Received: from int-mx2.corp.redhat.com (int-mx2.corp.redhat.com [172.16.27.26]) by mx2.redhat.com (8.13.8/8.13.8) with ESMTP id n1BLkpBc017587; Wed, 11 Feb 2009 16:46:51 -0500 Received: from ns3.rdu.redhat.com (ns3.rdu.redhat.com [10.11.255.199]) by int-mx2.corp.redhat.com (8.13.1/8.13.1) with ESMTP id n1BLkpxq004329; Wed, 11 Feb 2009 16:46:51 -0500 Received: from host0.dyn.jankratochvil.net (sebastian-int.corp.redhat.com [172.16.52.221]) by ns3.rdu.redhat.com (8.13.8/8.13.8) with ESMTP id n1BLkoWQ008536; Wed, 11 Feb 2009 16:46:51 -0500 Received: from host0.dyn.jankratochvil.net (localhost [127.0.0.1]) by host0.dyn.jankratochvil.net (8.14.3/8.14.3) with ESMTP id n1BLklaB024064; Wed, 11 Feb 2009 22:46:48 +0100 Received: (from jkratoch@localhost) by host0.dyn.jankratochvil.net (8.14.3/8.14.2/Submit) id n1BLklFT024061; Wed, 11 Feb 2009 22:46:47 +0100 Date: Wed, 11 Feb 2009 21:47:00 -0000 From: Jan Kratochvil To: Mark Kettenis Cc: drow@false.org, gdb-patches@sourceware.org Subject: Re: [patch] Fix `return' of long/long-long results with no debuginfo Message-ID: <20090211214646.GA22247@host0.dyn.jankratochvil.net> References: <20090211194511.GA9324@host0.dyn.jankratochvil.net> <200902112040.n1BKdxb3028188@brahms.sibelius.xs4all.nl> <20090211205045.GB9762@caradoc.them.org> <200902112122.n1BLMf8q000100@brahms.sibelius.xs4all.nl> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200902112122.n1BLMf8q000100@brahms.sibelius.xs4all.nl> User-Agent: Mutt/1.5.18 (2008-05-17) X-IsSubscribed: yes Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org X-SW-Source: 2009-02/txt/msg00263.txt.bz2 On Wed, 11 Feb 2009 22:22:41 +0100, Mark Kettenis wrote: > And I have a (somewhat) vague recollection that there are debug formats or > compilers where we cant distinguish between functions without a return type > and functions with "unknown" return type. As `struct type' for a function type GDB debuginfo readers always create by make_function_type and this function crashes on NULL `type' we can assume TYPE_TARGET_TYPE is never NULL. For DWARF it is clear both according to the standard 3.3.2 Subroutine and Entry Point Return Types If the subroutine or entry point is a function that returns a value, then its debugging information entry has a DW_AT_type attribute [...] and according to the code: dwarf2read.c:die_type(): /* A missing DW_AT_type represents a void type. */ Made the condition in the patch more explicit now. Removed the original TYPE_CODE_VOID code emulating the K&R behavior to more follow the modern ANSI/ISO C compilers refuse `return value;' in a void-returning function without having typeless/`int'-defaulted functions permitted. > From: Daniel Jacobowitz > > Everything we pick will be wrong some of the time, but IMO "long" is > > maximally useful. "long long" on 32-bit platforms is going to pick > > up garbage from the next register for int or void * returns. Garbage would be seen by caller of a `long long' returning callee where GDB would return something smaller (like `int'). As GDB writes (not "reads") the register it just will write zero to one excessive but unused register. This register must be callee-saved (not caller-saved) by the ABI supporting `long long' return type and GDB already asserts the return type is RETURN_VALUE_REGISTER_CONVENTION. If GDB would return only the `long' size there would be no way how to do the `return' command on a debuginfo-less function natively returning `long long'. > The right solution of course is to distribute libraries with debug > sumbols. But in cases where that is impossible, would it be an idea > to use the type of the return value expression given by the user > instead of int as a fallback? Any return value gets retyped to `long long' which is intepreted by the caller even as lower-sized types (not guaranteed by ABI but at lest checked now by new gdb.base/return-nodebug.exp). > In the case at hand, "return -1" will still fail, but it would be > possible to do return (void *)-1 to do what the user wanted. `return -1' is checked by gdb.base/return-nodebug.exp and it works. Thanks, Jan gdb/ 2009-02-11 Jan Kratochvil * stack.c (return_command): Default to the builtin_long_long or at least builtin_long types now. Assert TYPE_TARGET_TYPE of a typed function is non-NULL. Error on return value for void returning functions. gdb/testsuite/ 2009-02-11 Jan Kratochvil * gdb.base/return-nodebug.exp, gdb.base/return-nodebug.c: New. * gdb.base/return2.exp (try to return value in a void function): New. --- gdb/stack.c 11 Feb 2009 16:07:28 -0000 1.185 +++ gdb/stack.c 11 Feb 2009 21:39:57 -0000 @@ -1805,9 +1805,23 @@ return_command (char *retval_exp, int fr /* Cast return value to the return type of the function. Should the cast fail, this call throws an error. */ if (thisfun != NULL) - return_type = TYPE_TARGET_TYPE (SYMBOL_TYPE (thisfun)); - if (return_type == NULL) - return_type = builtin_type (get_frame_arch (thisframe))->builtin_int; + { + return_type = TYPE_TARGET_TYPE (SYMBOL_TYPE (thisfun)); + gdb_assert (return_type != NULL); + if (TYPE_CODE (return_type) == TYPE_CODE_VOID) + error (_("Function %s does not return any value"), + SYMBOL_PRINT_NAME (thisfun)); + } + else + { + return_type + = builtin_type (get_frame_arch (thisframe))->builtin_long_long; + + /* Hopefull each ABI returns at least `long' type in register. */ + if (using_struct_return (NULL, return_type)) + return_type + = builtin_type (get_frame_arch (thisframe))->builtin_long; + } CHECK_TYPEDEF (return_type); return_value = value_cast (return_type, return_value); @@ -1816,14 +1830,7 @@ return_command (char *retval_exp, int fr if (value_lazy (return_value)) value_fetch_lazy (return_value); - if (TYPE_CODE (return_type) == TYPE_CODE_VOID) - /* If the return-type is "void", don't try to find the - return-value's location. However, do still evaluate the - return expression so that, even when the expression result - is discarded, side effects such as "return i++" still - occur. */ - return_value = NULL; - else if (thisfun != NULL + if (thisfun != NULL && using_struct_return (SYMBOL_TYPE (thisfun), return_type)) { query_prefix = "\ --- /dev/null 1 Jan 1970 00:00:00 -0000 +++ gdb/testsuite/gdb.base/return-nodebug.c 11 Feb 2009 21:40:00 -0000 @@ -0,0 +1,40 @@ +/* This testcase is part of GDB, the GNU debugger. + + Copyright 2009 Free Software Foundation, Inc. + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . */ + +#include + +static TYPE +func (void) +{ + return (TYPE) -1; +} + +static void +marker (void) +{ +} + +int +main (void) +{ + printf ("result=" FORMAT "\n", CAST func ()); + + /* Cannot `next' with no debug info. */ + marker (); + + return 0; +} --- /dev/null 1 Jan 1970 00:00:00 -0000 +++ gdb/testsuite/gdb.base/return-nodebug.exp 11 Feb 2009 21:40:00 -0000 @@ -0,0 +1,52 @@ +# Copyright (C) 2009 Free Software Foundation, Inc. + +# This program is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 3 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see . + +proc do_test {typenospace} { + global pf_prefix + set old_prefix $pf_prefix + lappend pf_prefix "$typenospace:" + + if {[runto "func"]} { + # Test return from a function with no debug info (symbol; still it may + # have a minimal-symbol) if it does not crash. + gdb_test "return -1" "#0 .* main \\(.*" \ + "return from function with no debug info" \ + "Make selected stack frame return now\\? \\(y or n\\) " "y" + + # And if it returned the full width of the result. + gdb_test "adv marker" "\r\nresult=-1\r\n.* in marker \\(.*" \ + "full width of the returned result" + } + + set pf_prefix $old_prefix +} + +foreach case {{{char} %d (int)} \ + {{short} %d (int)} \ + {{int} %d} \ + {{long} %ld} \ + {{long long} %lld}} { + set type [lindex $case 0] + set format [lindex $case 1] + set cast [lindex $case 2] + + set typeesc [string map {{ } {\ }} $type] + set typenospace [string map {{ } -} $type] + + if {[prepare_for_testing return-nodebug.exp "return-nodebug-$typenospace" "return-nodebug.c" \ + [list "additional_flags=-DFORMAT=\"$format\" -DTYPE=$typeesc -DCAST=$cast"]] == 0} { + do_test $typenospace + } +} --- gdb/testsuite/gdb.base/return2.exp 3 Jan 2009 05:58:03 -0000 1.11 +++ gdb/testsuite/gdb.base/return2.exp 11 Feb 2009 21:40:00 -0000 @@ -70,6 +70,8 @@ proc return_void { } { "set break on void_func" gdb_test "continue" "Breakpoint.* void_func.*" \ "continue to void_func" + gdb_test "return 23" "Function void_func does not return any value" \ + "try to return value in a void function" send_gdb "return \n" gdb_expect { -re "Make void_func return now.*y or n. $" {