From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 104555 invoked by alias); 31 Aug 2019 13:02:24 -0000 Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Received: (qmail 103954 invoked by uid 89); 31 Aug 2019 13:02:24 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-10.8 required=5.0 tests=AWL,BAYES_00,GIT_PATCH_2,GIT_PATCH_3,SPF_PASS autolearn=ham version=3.3.1 spammy=Hi!, Hi X-HELO: mx1.suse.de Received: from mx2.suse.de (HELO mx1.suse.de) (195.135.220.15) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Sat, 31 Aug 2019 13:02:23 +0000 Received: from relay2.suse.de (unknown [195.135.220.254]) by mx1.suse.de (Postfix) with ESMTP id D2EC0B023; Sat, 31 Aug 2019 13:02:20 +0000 (UTC) Date: Sat, 31 Aug 2019 15:42:00 -0000 User-Agent: K-9 Mail for Android In-Reply-To: <20190831063003.GE2120@tucnak> References: <20190831063003.GE2120@tucnak> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Subject: Re: [PATCH] Fix free_lang_data on asm stmts (PR lto/91572) To: Jakub Jelinek CC: gcc-patches@gcc.gnu.org From: Richard Biener Message-ID: <6DE9EA1F-FDC4-4ADB-B3FB-213E2AD0D184@suse.de> X-SW-Source: 2019-08/txt/msg02131.txt.bz2 On August 31, 2019 8:30:03 AM GMT+02:00, Jakub Jelinek w= rote: >Hi! > >The following testcase ICEs, because on the inline asm LTO streaming >streams >the constraint strings ("g" in this case), including their type, but >the >fld type discovery doesn't see that type and so we end up streaming >const >char type turned into its own main variant. > >The strings for asm are in TREE_PURPOSE of the TREE_LIST args. >walk_tree doesn't walk TREE_PURPOSE though. Tried to change that, but >it >breaks way too much, tried to walk TREE_PURPOSE of TREE_LIST just for >the >fld walking (find_decls_types_r), but that doesn't work either, most of >the >TREE_PURPOSE we do not want to walk, usually it contains C++ default >arguments which fld clears. So, this directed patch walks the >TREE_PURPOSE >solely for the asm stmt arguments. > >Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? Ok, but I wonder if we can stream the constraint strings in a simpler way -= surely the type doesn't really matter? Why are they not identifier nodes?=20 Richard.=20 >2019-08-31 Jakub Jelinek > > PR lto/91572 > * tree.c (find_decls_types_in_node): Also walk TREE_PURPOSE of > GIMPLE_ASM TREE_LIST operands. > > * g++.dg/lto/pr91572_0.C: New test. > >--- gcc/tree.c.jj 2019-08-29 10:22:06.337702323 +0200 >+++ gcc/tree.c 2019-08-29 11:07:16.120107950 +0200 >@@ -6142,6 +6142,13 @@ find_decls_types_in_node (struct cgraph_ > { > tree arg =3D gimple_op (stmt, i); > find_decls_types (arg, fld); >+ /* find_decls_types doesn't walk TREE_PURPOSE of TREE_LISTs, >+ which we need for asm stmts. */ >+ if (arg >+ && TREE_CODE (arg) =3D=3D TREE_LIST >+ && TREE_PURPOSE (arg) >+ && gimple_code (stmt) =3D=3D GIMPLE_ASM) >+ find_decls_types (TREE_PURPOSE (arg), fld); > } > } > } >--- gcc/testsuite/g++.dg/lto/pr91572_0.C.jj 2019-08-28 >18:13:47.718349087 +0200 >+++ gcc/testsuite/g++.dg/lto/pr91572_0.C 2019-08-28 18:13:41.695436342 >+0200 >@@ -0,0 +1,12 @@ >+// PR lto/91572 >+// { dg-lto-do link } >+// { dg-lto-options { { -O -fPIC -flto } } } >+// { dg-require-effective-target shared } >+// { dg-require-effective-target fpic } >+// { dg-extra-ld-options "-shared" } >+ >+void foo (char); >+namespace N { >+ class A { A (); }; >+ A::A () { asm ("" : : "g" (0)); } >+} > > Jakub