X-Authentication-Warning: delorie.com: mail set sender to geda-user-bounces using -f X-Recipient: geda-user AT delorie DOT com Date: Mon, 14 Mar 2016 14:48:44 +0100 (CET) X-X-Sender: igor2 AT igor2priv To: geda-user AT delorie DOT com X-Debug: to=geda-user AT delorie DOT com from="gedau AT igor2 DOT repo DOT hu" From: gedau AT igor2 DOT repo DOT hu Subject: Re: [geda-user] pcb: more memory leaks (Bonus track: a short story on valgrinding the parser) In-Reply-To: Message-ID: References: User-Agent: Alpine 2.00 (DEB 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Reply-To: geda-user AT delorie DOT com Errors-To: nobody AT delorie DOT com X-Mailing-List: geda-user AT delorie DOT com X-Unsubscribes-To: listserv AT delorie DOT com Precedence: bulk 1. Easy - parse_y.y: rule pcbgroups: layer_group_string = $3; it's a dynamic allocated string in the lex that's stored in a global variable for later use. The variable then is never free'd. The reason is probably that another place layer_group_string gets assigned with a setting that should not be free'd. My solution was to strdup() the setting too, and always free the string after use. It's not super-efficient, but at least doesn't leak anymore. (r1286 - please ignore changes in the generated files) - there's still a leak by flex; it keeps a buffer allocated, but I don't see any trivial way to free it. As a bonus I figured how to trace back such leaks - a raw valgrind log for such a flex/bison string allocation is not very informative. For details, please refer to my devlog entry at http://repo.hu/projects/pcb-rnd/devlog/20160314_valgrind_flex.html Regards, Igor2 P.S. with the batch hid, for the simple load-and-exit test there are only 4 leaks left in pcb-rnd (77 bytes total). Two of these are pcb-rnd specific and will be easy to fix. The other two are dlopen() and flex leaks which I probably won't be able to fix. Next tests will be with some user actions added.