From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 26213 invoked by alias); 26 Sep 2003 18:39:25 -0000 Mailing-List: contact gdb-help@sources.redhat.com; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-owner@sources.redhat.com Received: (qmail 26171 invoked from network); 26 Sep 2003 18:39:23 -0000 Received: from unknown (HELO nevyn.them.org) (66.93.172.17) by sources.redhat.com with SMTP; 26 Sep 2003 18:39:23 -0000 Received: from drow by nevyn.them.org with local (Exim 4.22 #1 (Debian)) id 1A2xUw-0002Ke-PQ; Fri, 26 Sep 2003 14:39:22 -0400 Date: Fri, 26 Sep 2003 18:39:00 -0000 From: Daniel Jacobowitz To: Andrew Greenlaw Cc: gdb@sources.redhat.com Subject: Re: "Cannot access memory at address 0x175f80" Message-ID: <20030926183922.GA8920@nevyn.them.org> Mail-Followup-To: Andrew Greenlaw , gdb@sources.redhat.com References: <3F747A00.3090302@agere.com> <20030926174934.GB23276@nevyn.them.org> <3F748295.9090404@agere.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <3F748295.9090404@agere.com> User-Agent: Mutt/1.5.1i X-SW-Source: 2003-09/txt/msg00328.txt.bz2 On Fri, Sep 26, 2003 at 02:16:53PM -0400, Andrew Greenlaw wrote: > I was specifying add-symbol-file, because when I start "orange", > none of the symbols from "apple" are accessible. In DDD, the source > code for Apple's files don't even appear in the "open source" menu. The > only 2 ways to make those files appear are 1) to run the program once, > then set breakpoints, then re-run, or 2) use the add-symbol-file > command. I admit I wasn't using it in an informed manner, but it worked > for the other symbols. > > I tried a little experiment: I started gdb, used the > add-symbol-file command, then entered "break nc_signal_raised". It said > "Cannot access memory at address 0x175f9c". I then exited gdb, started > over, and this time, instead of add-symbol-file, ran the program. when > the program exitted, I typed "break nc_signal_raised", It said > "Breakpoint 1 at 0x40fb2f9f: file /sopt/ldv-dev/tools/src/main.cc, line > 8.". So that was part of the problem. > > Now the bad news. If I try to re-run, this particular breakpoint > can't be inserted. After that encouraging message above, I hit the > "run" button in ddd, and got: > > (gdb) run snap1 > (no debugging symbols found)...Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > ncsim: 05.00-s005: (c) Copyright 1995-2003 Cadence Design Systems, Inc. > ncsim: *W,DLNOHV: Unable to find an 'hdl.var' file to load in. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > Error in re-setting breakpoint 1: > Function "nc_signal_raised" not defined. > > Any Ideas? Sure, it's noisy, but does the breakpoint eventually get defined? If not, set stop-on-solib-events to 1 and then hit continue until info shared reports that module loaded. Then enable the breakpoint. > > Daniel Jacobowitz wrote: > > >On Fri, Sep 26, 2003 at 01:40:16PM -0400, Andrew Greenlaw wrote: > > > > > >>Hi. I'm debugging a C++ tool (called apple) that's loaded as a > >>dynamic library into a C - or C++ program (called orange) . Apple is > >>compiled with g++ 3.3.1, binutils 2.14, and debugging symbols are > >>enabled. Orange is a big unknown (meaning: I don't know how it was > >>compiled) and it has no debugging symbols. The gdb version is 5.3 > >> > >> Here are the flags used to compile Apple: > >># -gdwarf-2 -g3 used to enable stepping through macro execution under > >>gdb 5.3 The explanation's under gdb 5.3 release notes. > >> Compile:g++ -std=c++98 -Wall -DLINUX -gdwarf-2 -g3 -D__USE_GNU > >>-D_GNU_SOURCE -fPIC file_name.cpp > >> > >> > >> Link: > >> g++ -shared $(LIBS) $(OBJSCHEF) $(PLIOBJSVCS_PLI) > >>InterfaceObjectVCS_PLI.o -o $@ -lc gdwarf-2 -g3 -fPIC > >>g++ -MD -std=c++98 -shared -lc -gdwarf-2 -g3 -fPIC -o > >>libapple.so > >> > >> When I go to debug apple, I run "gdb ", then use the > >>"add-symbol-file -readnow" gdb comman to load > >>the symbols from apple. From there, I can set breakpoints in C++ class > >>methods, no problem. But, there is 1 function (so, non-OO code), where > >>if I try to set a breakpoint, I get the following: > >> > >> (gdb) break nc_signal_raised > >>Cannot access memory at address 0x175f80 > >> > >> If I do an "nm" on the library, I get: > >> 00175f80 T nc_signal_raised > >> > >> Indicating that the address read by gdb is correct. > >> > >> > > > >Unlikely, since the library is not loaded at a base address of 0. > > > > > > > >> On a related, but less important note, when use ddd to debug & go to > >>a source file, I always get: "" is at address 0x10f5f0 > >> but contains no code. > >> > >> And yet I can set breakpoints or step through the code. What's > >>going on? > >> > >> Any help you can offer will be appreciated. I've been working on > >>this for 2 weeks, read every posting or piece of documentation I can > >>find. I'm at my wits' end! > >> > >> > > > >Why are you using add-symbol-file? Is the loader not done as a dlopen, > >i.e. are you dealing with something that has its own dynamic loader? > >If it's dlopen'd, gdb should automatically handle it. > > > >It seems unlikely that add-symbol-file without specifying a text offset > >is right, too. > > > > > > > > -- > Andrew Greenlaw, Advanced Verification Group > Agere Systems of Ottawa, ON. Intranet page: > http://ottawa/~andrewg/ > E-mail: andrewg@agere.com, Phone: (613)768-8738, Fax: > (768)768-8710 > **** > "A mathematician is a device for turning coffee into theorems." > - Paul Erdos > **** > > > > -- Daniel Jacobowitz MontaVista Software Debian GNU/Linux Developer