[PATCH] MCJIT MachO debugging support

[Lang Hames]

Hi Keno,

ELF is broken too, and I'll be looking at fixing that in the future. The
impact on MachO is higher as MachO relies on the section address info for
relocation, which ELF does not.

I would love to fix this by changing the debugger interface. I was planning
to send an email out about that this afternoon. Do you work on the GDB side
of this interface?

Cheers,
Lang.


On Tue, Jul 15, 2014 at 1:40 PM, Keno Fischer <kfischer at college.harvard.edu>
wrote:

> The ELF backend does the exact same thing here. Should we rip that out
> too? Otherwise, could we maybe selectively enable it for the MCJIT case? I
> guess the real problem here is the debugger interface, which requires this
> kind of thing. Might be good to sit down very soon and think about if
> there's a better interface to the debugger we should expose. It would be
> good to do that sooner rather than later.
>
>
> On Tue, Jul 15, 2014 at 12:54 PM, Lang Hames <lhames at gmail.com> wrote:
>
>> Hi Keno,
>>
>> I've reverted this in r213086 and filed
>> http://llvm.org/bugs/show_bug.cgi?id=20309 to track re-enabling support.
>>
>> Cheers,
>> Lang.
>>
>>
>>
>> On Tue, Jul 15, 2014 at 11:30 AM, Lang Hames <lhames at gmail.com> wrote:
>>
>>> Hi Keno,
>>>
>>> Sorry - I'm going to have to back this out: It's breaking llvm-rtdyld
>>> due to the read-only MemoryBuffer issue. It also wreaks havoc with
>>> RuntimeDyldMachO in general, since we can no longer code to the
>>> MachOObjectFile API (section address references are common, and now these
>>> all need to go via the ObjectImage).
>>>
>>> Once the RuntimeDyldMachO cleanup is in (hopefully today) I'll return to
>>> this - I think I can adapt your patch to build a solution that will do the
>>> right thing here.
>>>
>>> Cheers,
>>> Lang.
>>>
>>>
>>>
>>> On Tue, Jun 24, 2014 at 5:29 PM, Lang Hames <lhames at gmail.com> wrote:
>>>
>>>> Committed in r211652. Thanks Keno!
>>>>
>>>> http://reviews.llvm.org/D4005
>>>>
>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140715/04bad734/attachment.html>