From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 27432 invoked by alias); 8 Aug 2007 21:46:55 -0000 Received: (qmail 27404 invoked by uid 22791); 8 Aug 2007 21:46:55 -0000 X-Spam-Check-By: sourceware.org Received: from mail.codesourcery.com (HELO mail.codesourcery.com) (65.74.133.4) by sourceware.org (qpsmtpd/0.31) with ESMTP; Wed, 08 Aug 2007 21:46:50 +0000 Received: (qmail 10323 invoked from network); 8 Aug 2007 21:46:49 -0000 Received: from unknown (HELO localhost) (jimb@127.0.0.2) by mail.codesourcery.com with ESMTPA; 8 Aug 2007 21:46:49 -0000 To: msnyder@sonic.net Cc: gdb-patches@sourceware.org Subject: Re: [PATCH] solib_open, memory leak References: <21906.12.7.175.2.1186283848.squirrel@webmail.sonic.net> <18682.12.7.175.2.1186597784.squirrel@webmail.sonic.net> From: Jim Blandy Date: Wed, 08 Aug 2007 21:46:00 -0000 In-Reply-To: <18682.12.7.175.2.1186597784.squirrel@webmail.sonic.net> (msnyder@sonic.net's message of "Wed, 8 Aug 2007 11:29:44 -0700 (PDT)") Message-ID: User-Agent: Gnus/5.11 (Gnus v5.11) Emacs/22.0.50 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii 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: 2007-08/txt/msg00160.txt.bz2 msnyder@sonic.net writes: >> I think the invariant should be that, when found_file becomes >= 0, >> then temp_pathname is malloc'd, and not before. The 'openp' clauses >> will preserve that. So I think you need: >> >> if (found_file >= 0) >> temp_pathname = xstrdup (temp_pathname); >> >> after the 'open'. >> >> And then there's no need for the xstrdup at the bottom; just return >> temp_pathname, or xfree it if the caller doesn't want it. > > That sounds great, except for one thing (which I had also not addressed). > > If openp receives an xmalloc'd temp_pathname, it will clobber it > without freeing it. > > > I'm thinking that openp will never use the buffer, only the pointer. Right. > So rather than what you suggest, what if we AVOID xmalloc before > openp and just make sure that we pass openp a null pointer or a pointer > to memory that does not need to be freed (such as alloca). > > Umm, thus: You're right that temp_pathname needs to be NULL if we haven't found something, so that the *found_pathname code at the end works right. The code clearly deserves a comment like: /* We try to find the library in various ways. After each attempt, either found_file >= 0 and temp_pathname is a malloc'd string, or found_file < 0 and temp_pathname does not point to storage that needs to be freed. */ In that light, it seems clearer to me to just put: if (found_file < 0) temp_pathname = NULL; else temp_pathname = xstrdup (temp_pathname); immediately after the open, to make it obvious that the rule is followed there. The stuff about making in_pathname relative doesn't affect whether anything is found or the allocatedness of temp_pathname, so it should be left alone. > *************** solib_open (char *in_pathname, char **fo > *** 224,229 **** > --- 235,247 ---- > > /* Done. If not found, tough luck. Return found_file and > (optionally) found_pathname. */ > + if (temp_pathname) > + { > + if (found_pathname) > + *found_pathname = temp_pathname; > + else > + xfree (temp_pathname); > + } > if (found_pathname != NULL && temp_pathname != NULL) > *found_pathname = xstrdup (temp_pathname); > return found_file; Don't you want those new lines to replace the two that followed them, not just precede them?