From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 9704 invoked by alias); 14 Feb 2007 16:29:22 -0000 Received: (qmail 9697 invoked by uid 22791); 14 Feb 2007 16:29:22 -0000 X-Spam-Status: No, hits=-0.9 required=5.0 tests=AWL,BAYES_50,SUBJ_HAS_UNIQ_ID X-Spam-Check-By: sourceware.org Received: from agminet01.oracle.com (HELO agminet01.oracle.com) (141.146.126.228) by sourceware.org (qpsmtpd/0.31) with ESMTP; Wed, 14 Feb 2007 16:29:10 +0000 Received: from rgmsgw02.us.oracle.com (rgmsgw02.us.oracle.com [138.1.186.52]) by agminet01.oracle.com (Switch-3.2.4/Switch-3.1.7) with ESMTP id l1EGT7nQ003076 for ; Wed, 14 Feb 2007 10:29:07 -0600 Received: from [10.8.0.10] (ca-server1.us.oracle.com [139.185.48.5]) by rgmsgw02.us.oracle.com (Switch-3.2.4/Switch-3.2.4) with ESMTP id l1EGSsmK022434; Wed, 14 Feb 2007 09:28:55 -0700 Message-ID: <45D33830.3070800@oracle.com> Date: Wed, 14 Feb 2007 16:29:00 -0000 From: Elena Zannoni Organization: Oracle USA Inc. User-Agent: Thunderbird 1.5.0.7 (X11/20060913) MIME-Version: 1.0 To: frysk Subject: notes ui call 20070214 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Brightmail-Tracker: AAAAAQAAAAI= X-Brightmail-Tracker: AAAAAQAAAAI= X-Whitelist: TRUE X-Whitelist: TRUE X-IsSubscribed: yes Mailing-List: contact frysk-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Post: List-Help: , Sender: frysk-owner@sourceware.org X-SW-Source: 2007-q1/txt/msg00125.txt.bz2 In case these are helpful, here are my notes for today's call. Attendees: RH: Andrew, Phil, Rich, Chris, Sami, Nurdin, Mike, Tim, Mark, Jonathan Oracle: Kris, Elena HPDF interface, with fhpd. -- Tim and Stan Demonstrated command set with attach, go, halt, where, help, list Tim: stop the world or stop individual threads? Right now it's stop the world. But underlying mechanism to do stop each thread is there. Tab completion: demoed go is different from continue in hpdf. continue is blocking the hpd interface, go doesn't. I.e. go is async, continue is sync. breakpoints: can set breakpoints, and run (there is a bug, though that Tim has a fix for, not checked in yet). Break syntax right now is break needs refining. print command can display arrays, array elements and classes Can also print in different formats. print -format [x|o] Help is totally broken Can do "what " displays the type of the variable or function disassembler: not hooked up to the hpd yet. Works in the source window. Not demoed. display of registers: there needs to be a refactoring. Not demoed. volatile attribute can make a variable stick around even with optimization. how about logging, should that be an option on start up of fhpd. compatibility with gdb command set. it's needed, maybe a set of aliases. Custom observers -- Sami cagney: use "catch" instead of "Print" in the actions. To avoid printing the whole backtrace for instance, since it can be too long. Just show that the event has happened and stop there. Fork observer has 2 callback functions: in one the parent is paused, in the other the child is paused. Picking an action on either thread, the thread will be invoking the proper callback. Some actions take arguments, like "add observer". Enter args in the box to the right of the action selection box. There is an autocompletion feature that is now broken, on the args. no way to tell if all filters are valid or not for a given action if filters and actions have been applied. plus/minus buttons are confusing. Minus on the first one: what does it do? unclear. Is order of actions important? How common is it to create custom observers? Frysk developers use them. Custom written events are needed. Written using code. Filters should be a short list Interactions between actions and the radio buttons at the bottom. Users could construct contradictory events/actions. Open issue: How to best bind events to threads. One event can apply to many threads Conference call issues: 1. Wednesday 9:30AM seems to be better than Thursday 9:30AM. All agree to move call to Wed. 2. Request to open up the RH only conf call to all. For general status of the project, who works on what, progress reports, and dependencies/issue with other components (kernel, gcc, libunwind, etc) RH has things that cannot be discussed openly, will need to figure out how to avoid having 3 meetings. Elena's suggestion is to use the Wed call, and shorten the GUI part.