Re: GDB 13 release -- 2023-01-21 Update

[Luis Machado <luis.machado@arm.com>]

Hi Joel,

On 1/31/23 07:03, Joel Brobecker wrote:
> Hi Luis,
> 
> On Mon, Jan 30, 2023 at 11:23:06AM +0000, Luis Machado wrote:
>> On 1/29/23 11:52, Joel Brobecker wrote:
>>>>>>>       - With that said, the patch appears to simply add a cache,
>>>>>>>         so the logic of it all doesn't appear to be extremely
>>>>>>>         complicated. So I would rate the risk to be low.
>>>>>>
>>>>>> That's what it seems to me.
>>>>>>
>>>>>> I think another round of testing would be a good step to make sure there are no hidden bugs.
>>>>> My understanding is that the problem was making debugging borderline
>>>>> impossible.  Or at least, really, really unpleasant.  So I think it
>>>>> qualifies as a bug fix, and that the benefit is worth the risk.
>>>>
>>>> Right. Past a short number of frames, unwinding got really really slow.
>>>
>>> The part that I haven't been clear on is whether this affected everyone
>>> on Arm-32bit, or everyone on both Arm and AArch64, or just a subset
>>> of the users?
>>
>> Just a subset.
>>
>> It affects 32-bit m-profile Arm targets that report the additional stack pointers.  This is either
>> the org.gnu.gdb.arm.m-system feature or the org.gnu.gdb.arm.secext feature.
>>
>> I'm fairly sure these features are only reported by emulators and bare-metal targets.
>>
>> Now, a 64-bit gdb can debug 32-bit Arm as well, so you could have a 64-bit gdb running into this too
>> if the target is a 32-bit m-profile Arm. But the target is still 32-bit Arm.
>>
>> This doesn't affect AArch64 at all.
> 
> Thanks for the extra effort explaining the impact. This is much clear
> for me, now. With your permission, I'd like to copy the text above
> as a comment in the PR. Would that be OK?> 

Yes, that's fine. Let me know if you want me to do it.

>>> Regardless of the above, I agree we can backport. To avoid confusion,
>>> can I leave this to you, Luis, to confirm that the two patches I identified
>>> previously are the only two patches that need to be backported, and
>>> can you backport those for us, with a round of testing if you haven't
>>> done so already, just to double-check?
>>
>> Sure. I gave it a try with master and the test results look the same with or without the patches.
>>
>> Unfortunately the 32-bit Arm test results are not too clean as there is quite a bit of noise from
>> failing watchpoints tests.
>>
>> I'll try with GDB 13 just to be sure.
> 
> Thank you. Once you've finished testing, whatever that might be,
> you can go ahead and push the patches. As disussed earlier, we all
> agree the risk of bad impact is very low.
> 
> Let me know when this is done. It's starting to look like we might
> have a pre-release created this weekend! Maybe earlier, if I can
> carve some time out of my work schedule.
> 

I ran checks with the gdb 13 branch and the results look the same for a patched gdb compared to
an unpatched one.

I'll cherry-pick both commits to gdb 13 once sourceware is back up.

Thanks for the patience in getting this addressed.

>>> For the record, those patches were:
>>>
>>>       commit d72ba177c85f2ad18d0dcabdd8844532c9acb819
>>>       Author: Torbj�rn SVENSSON <torbjorn.svensson@foss.st.com>
>>>       Date:   Thu Nov 17 12:17:53 2022 +0100
>>>       Subject: gdb: dwarf2 generic implementation for caching function data
>>>
>>> ... and ...
>>>
>>>       commit 5cf11483141a58314834653003e49709b47822d5
>>>       Author: Torbj�rn SVENSSON <torbjorn.svensson@foss.st.com>
>>>       Date:   Thu Nov 17 12:18:20 2022 +0100
>>>       Subject: gdb/arm: Use new dwarf2 function cache
>>>
>>> Thank you!
>>
>