[RFC] [gdb/breakpoints] Use gdbarch_addr_bits_remove in breakpoint_address_match

[Luis Machado]

On 9/23/24 09:30, Tom de Vries wrote:
> On 8/30/24 10:53, Luis Machado wrote:
>> It is difficult to have a proper solution to determining thumbness on 32-bit arm. Given
>> the improvement, I don't see a problem with this patch, so unless others have objections...
>>
>> Approved-By: Luis Machado<luis.machado@arm.com>
> 
> Hi Luis,
> 
> thanks for the review.
> 
> Just to re-iterate in more detail what I mentioned in a recent off-list conversation.
> 
> I haven't committed this patch, because while analyzing another failure on arm-linux I came to realize that there are other places in gdb that compare (breakpoint) addresses that do not use breakpoint_address_match.
> 
> Fixing all of those spots would be some work, but I'm not sure it's a good idea to begin with, because it would also be difficult to maintain if this only gets us in trouble on one target, in absence of minimal symbols, which happens on a limited amount of distros.
> 
> So until I have re-convinced myself that this is the way to go, I'm postponing committing this.
> 
> Thanks,
> - Tom

Thanks for the heads-up. There might be ways to improve the situation somewhat, but unfortunately reliability depends on the presence of the thumb mapping symbols.