[lldb-dev] "bad_alloc" crash after downloading dependent modules.

[Pavel Labath]

Hello Aidan,

thanks for the information. We have identified the cause of the
problem to be an incorrect value passed through the JIT-debugger
interface <https://sourceware.org/gdb/onlinedocs/gdb/Declarations.html#Declarations>.
Specifically, the jit_code_entry:: symfile_size value is wrong. It
looks like it was written as a 32-bit value, whereas the "spec" says
the value should be 64-bit, regardless of the architecture.

Now, obviously wrong value should not crash the debugger, and we will
fix this. However, the main problem is that the wrong value is being
supplied over the interface. As I understand it, you (your team) are
in charge of the code which generates these structures. Could you
verify that this field is being set correctly?

cheers,
pl


On 11 June 2015 at 05:04, Aidan Dodds <aidan at codeplay.com> wrote:
> Hi Pavel,
>
> Thanks for getting back to me.
> This seems to be a regression as this was working for us a month or so ago.
>
> The device I am using is a new Nexus 6 phone, flashed with a recent build of
> AOSP master.
> I have attached a zip file which contains the APK I was debugging, which is
> just the BasicRenderScript example that comes with android studio.
> I have also attached the adb bug report that you had asked for, as well a a
> transcript of the crash with some lldb logs enabled.
>
> My LLDB and LLDBServer were build from the same checkout, at the following
> commits from the GIT Mirror:
>
> LLVM
> commit 544d686bc0d6e10f483106829e4f7a69346d4b26
> Author: Igor Laevsky <igmyrj at gmail.com>
> Date:   Wed Jun 10 12:31:53 2015 +0000
>
> CLANG
> commit c024327365be5c472f3a1afd222161f6aa4ee2d6
> Author: Alexander Musman <alexander.musman at gmail.com>
> Date:   Wed Jun 10 11:20:26 2015 +0000
>
> LLDB
> commit efc91053b816d4a2d020c47287fb02888d5f26e0
> Author: Mohit K. Bhakkad <mohit.bhakkad at gmail.com>
> Date:   Wed Jun 10 10:02:21 2015 +0000
>
> If I can provide any more information or help you out in any way just let me
> know.
>
> Thanks,
> Aidan
>
>
> On 10/06/2015 18:18, Pavel Labath wrote:
>>
>> Hello Aidan,
>>
>> Thanks for the report. android-arm is not as stable as we would like
>> it to be, but we are actively working on changing that. Could you
>> provide us with more information about your setup:
>> - which android version/device are you encountering this on (try
>> running "adb bugreport")?
>> - which application are you attaching this to (if possible, send me an
>> apk)
>>
>> thanks,
>> pl
>>
>> ps: these things can be big, so feel free to send it to me privately
>>
>>
>> On 10 June 2015 at 08:12, Aidan Dodds <aidan at codeplay.com> wrote:
>>>
>>> Hi Folks,
>>>
>>> I have been seeing a crash when connecting lldb tip to a remote arm
>>> lldb-server platform running on android.
>>> The crash occurs directly after all the remote modules have been
>>> downloaded.
>>> I have attached a transcript of a session which shows the error and
>>> displays
>>> a backtrace after the crash.
>>>
>>> The same crash can be reproduced on both linux and windows.
>>>
>>> Might I be doing something wrong with my setup / build?
>>> Is 32bit Arm + Android a stable target for lldb / lldb-server or is this
>>> somewhat unstable?
>>>
>>> I should point out that if I connect to lldb-server in gdbserver mode
>>> rather
>>> then platform this problem doesn't present itself.
>>> That's perhaps because in that case the platform defaults to remote-linux
>>> rather then remote-android.
>>>
>>> Thanks,
>>> Aidan
>>>
>>> _______________________________________________
>>> lldb-dev mailing list
>>> lldb-dev at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev
>>>
>