From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 2987 invoked by alias); 25 Jan 2015 19:32:45 -0000 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 Received: (qmail 2877 invoked by uid 89); 25 Jan 2015 19:32:19 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-1.0 required=5.0 tests=AWL,FREEMAIL_ENVFROM_END_DIGIT,FREEMAIL_FROM,RCVD_IN_DNSWL_LOW,SPF_PASS autolearn=ham version=3.3.2 X-HELO: mail-pd0-f177.google.com Received: from mail-pd0-f177.google.com (HELO mail-pd0-f177.google.com) (209.85.192.177) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with (AES128-SHA encrypted) ESMTPS; Sun, 25 Jan 2015 19:32:08 +0000 Received: by mail-pd0-f177.google.com with SMTP id y13so7931012pdi.8 for ; Sun, 25 Jan 2015 11:31:58 -0800 (PST) X-Received: by 10.70.91.201 with SMTP id cg9mr28697031pdb.57.1422214317997; Sun, 25 Jan 2015 11:31:57 -0800 (PST) Received: from seba.sebabeach.org.gmail.com (173-13-178-53-sfba.hfc.comcastbusiness.net. [173.13.178.53]) by mx.google.com with ESMTPSA id xw1sm8007230pac.47.2015.01.25.11.31.56 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Sun, 25 Jan 2015 11:31:57 -0800 (PST) From: Doug Evans To: gdb-patches@sourceware.org, gaius@glam.ac.uk Subject: [PATCH 0/5] Remove struct main_type.vptr_{fieldno,basetype} Date: Sun, 25 Jan 2015 19:56:00 -0000 Message-ID: MIME-Version: 1.0 Content-Type: text/plain X-IsSubscribed: yes X-SW-Source: 2015-01/txt/msg00683.txt.bz2 Hi. Anytime you can reduce the size of struct main_type that's A Good Thing. Alas in this case my motivation, besides more general cleanup (will it never end?), is that I may wish to add something to struct main_type. Now I can, if I need to. There are two fields in struct main_type that really shouldn't be there: vptr_fieldno, and vptr_basetype. Every type has to pay the price in memory usage for their presence. The proper place for them is in the type_specific field (in this case cplus_stuff) which exists for this purpose. Things are a bit complicated because vptr_basetype is overloaded to record the type of "this" for TYPE_CODE_{METHODPTR,MEMBERPTR,METHOD}, and is accessed through the macro TYPE_DOMAIN_TYPE. There is also the wrinkle, discussed earlier, that m2-typeprint.c uses TYPE_DOMAIN_TYPE. I've replaced that with TYPE_TARGET_TYPE and left a FIXME there (pre-discussed with Gaius). This is done in 4/5. The series has been tested on amd64-linux with dwarf and stabs. It needs more testing, eg., on other file formats (e.g., xcoff). And it could use another set of eyes. One thing I did was rename TYPE_DOMAIN_TYPE to TYPE_SELF_TYPE. My mind is far too used to associating the word "domain" with VAR_DOMAIN, STRUCT_DOMAIN, etc. A better name would be nice. I went with "self" because TYPE_THIS_TYPE was a bit confusing. Another alternative is TYPE_TYPE_OF_THIS, but TYPE_SELF_TYPE "works for me". Internally, some of the debug info readers still use "domain", which is fine by me, and I've left those alone. 1/5 - Finish TYPE_SPECIFIC_FIELD handling in copy_type_recursive 2/5 - Tighten up handling of types passed in gnu-v3-abi.c ["info vtbl my_int" will currently call INIT_CPLUS_SPECIFIC on that int!] 3/5 - Rename TYPE_DOMAIN_TYPE to TYPE_SELF_TYPE 4/5 - Create TYPE_SPECIFIC_SELF_TYPE and have TYPE_CODE_{METHODPTR, MEMBERPTR,METHOD} use it. 5/5 - Finally, move vptr_fieldno,vptr_basetype into cplus_struct_type.