[yocto] Can't enable reverse debugging

Mon Dec 12 13:48:14 PST 2016

> -----Original Message-----
> From: Khem Raj [mailto:raj.khem at gmail.com]
> Sent: Monday, December 12, 2016 2:54 PM
> To: Bryan Evenson <bevenson at melinkcorp.com>
> Cc: yocto at yoctoproject.org
> Subject: Re: [yocto] Can't enable reverse debugging
> 
> 
> > On Dec 12, 2016, at 11:12 AM, Bryan Evenson
> <bevenson at melinkcorp.com> wrote:
> >
> > Khem,
> >
> >> -----Original Message-----
> >> From: Khem Raj [mailto:raj.khem at gmail.com]
> >> Sent: Monday, December 12, 2016 12:48 PM
> >> To: Bryan Evenson <bevenson at melinkcorp.com>
> >> Cc: yocto at yoctoproject.org
> >> Subject: Re: [yocto] Can't enable reverse debugging
> >>
> >>
> >>> On Dec 12, 2016, at 6:59 AM, Bryan Evenson
> <bevenson at melinkcorp.com>
> >> wrote:
> >>>
> >>> Khem,
> >>>
> >>>> -----Original Message-----
> >>>> From: Khem Raj [mailto:raj.khem at gmail.com]
> >>>> Sent: Saturday, December 10, 2016 4:26 PM
> >>>> To: Bryan Evenson <bevenson at melinkcorp.com>
> >>>> Cc: yocto at yoctoproject.org
> >>>> Subject: Re: [yocto] Can't enable reverse debugging
> >>>>
> >>>> On Fri, Dec 9, 2016 at 8:43 AM, Bryan Evenson
> >>>> <bevenson at melinkcorp.com>
> >>>> wrote:
> >>>>> I tried a few other things, see below.
> >>>>>
> >>>>>> -----Original Message-----
> >>>>>> From: yocto-bounces at yoctoproject.org [mailto:yocto-
> >>>>>> bounces at yoctoproject.org] On Behalf Of Bryan Evenson
> >>>>>> Sent: Thursday, December 08, 2016 2:50 PM
> >>>>>> To: yocto at yoctoproject.org
> >>>>>> Subject: [yocto] Can't enable reverse debugging
> >>>>>>
> >>>>>> I'm on poky/dizzy and I'm getting a segementation fault in my
> >>>>>> application that I think may be related to a buffer overflow.
> >>>>>> I'm using the Eclipse
> >>>> plug-in
> >>>>>> (on Eclipse Kepler) to debug on the hardware with GDB.  I'm
> >>>>>> trying to
> >>>> enable
> >>>>>> reverse debugging but I have not had any success and I'm looking
> >>>>>> for
> >>>> some
> >>>>>> help.
> >>>>>>
> >>>>>> My target processor is an SAM9G25 (ARM926EJ-S core).  I have been
> >>>>>> able
> >>>> to
> >>>>>> debug through Eclipse on the target hardware without issue.  I
> >>>>>> tried
> >>>> enabling
> >>>>>> Reverse Debugging by going to Window->Customize Perspective...-
> >>>>>>> Command Groups Availability and selecting "Reverse Debugging".
> >>>>>>> Now
> >>>>>> there is a "Reverse Toggle" button on my toolbar, but it is
> >>>>>> always grayed
> >>>> out
> >>>>>> whether I have a debug session running or not.  I have also tried
> >>>>>> entering
> >>>> the
> >>>>>> command "record" from the GDB console after my program breaks
> at
> >>>> main.
> >>>>>> When I run the program, I get the error message "Process record:
> >>>>>> failed
> >>>> to
> >>>>>> record execution log".
> >>>>>
> >>>>> I tried first running my application, pausing the debugger and
> >>>>> then sending
> >>>> the "record" command.  When I continued debugging, I got the
> >>>> following GDB error:
> >>>>>
> >>>>> ....x86_64-nativesdk-pokysdk-linux/gdb-cross-canadian-arm/7.7.1-r0
> >>>>> /g
> >>>>> db-
> >>>> 7.7.1/gdb/utils.c:1073: internal-error: virtual memory exhausted.
> >>>>> A problem internal to GDB has been detected, further debugging may
> >>>>> prove unreliable.
> >>>>>
> >>>>> I did some looking around and I couldn't see any reasonable answer
> >>>>> to why
> >>>> this would happen.  I also just noticed that for the Debug
> >>>> Configurations under Eclipse, the debugger settings tab doesn't
> >>>> have "Enable Reverse Debugging at startup" option listed for the
> >>>> C/C++ Remote Application debugging type.  Is this option just not
> >>>> available to me
> >> with my environment?
> >>>>
> >>>> This could mean many things but one area I know is troublesome is
> >>>> python support. Sometimes there are issues between fundamental
> >> types,
> >>>> can you check if the error happens when gdb is running some
> >>>> pythonic stuff.
> >>>
> >>> Could you explain how to do this?  I don't use Python so I'm not
> >>> sure if I'm
> >> doing a valid test.  I tried to debug a Python script with GDB
> >> through Eclipse and I wasn't successful.  I then went through the
> >> command line and was able to remotely run a simple Python script
> >> (print "Hello, Python" once per
> >> second) through GDB.  If I issued a "record" command through GDB
> >> first, then when I tried to debug the Python script I'd immediately
> >> get a segmentation fault.
> >>>
> >>> Is there something specific you'd like me to try?  If you think the
> >>> Python
> >> support may be the issue, is there something I should be verifying is
> >> present or is at a certain version?
> >>>
> >>
> >> Disable python support in gdb and compile it e.g.
> >
> > I rebuilt gdb.  I used the --configure option to verify that my previous gdb
> did have the "--with-python" option present and that my newly built gdb did
> not have that option anymore.  I no longer can produce the " internal-error:
> virtual memory exhausted" error, but whenever I try to record I still get the
> "Process record: failed to record execution log" when I run the application.  If
> it helps, here's the configuration output from GDB:
> >
> > This GDB was configured as follows:
> >   configure --host=x86_64-linux --target=arm-poky-linux-gnueabi
> >             --with-auto-load-dir=$debugdir:$datadir/auto-load
> >             --with-auto-load-safe-path=$debugdir:$datadir/auto-load
> >             --with-expat
> >             --with-gdb-datadir=~/poky/poky-build/tmp/sysroots/x86_64-
> linux/usr/share/gdb-arm926ejste-poky-linux-gnueabi/gdb (relocatable)
> >             --with-jit-reader-dir=~/poky/poky-build/tmp/sysroots/x86_64-
> linux/usr/lib/arm-poky-linux-gnueabi/gdb (relocatable)
> >             --without-libunwind-ia64
> >             --without-lzma
> >             --with-separate-debug-dir=~/poky/poky-
> build/tmp/sysroots/x86_64-linux/usr/lib/arm-poky-linux-gnueabi/debug
> (relocatable)
> >             --with-zlib
> >             --without-babeltrace
> >
> > Any additional suggestions would be appreciated.
> 
> Can you build an application with out splitting the debug info. May be have
> toolchain on target and then compile a program natively and see if that works

For the application I am debugging through Eclipse I am using a self-built SDK to build the application.  I verified that the debugging info has not been split out from the executable that I'm using for testing.  The target I'm using only has 256MB of flash and 128MB of RAM, so I'm not sure if I'd be able to build the application natively.

Bryan