Re: gdb 8.3.1 truncated register in remote g packet

On 11/2/19 9:07 PM, Reinoud Koornstra wrote:
> 
> 
> 
>     I'm not sure why an arm-targeted gdb would not complain about loading a
>     x86-target vmlinux image. That sounds off.
> 
> The vmlinux was compiled for arm, the thing that surprised me that the 
> x64 gdb didn't complain about it. The arm gdb didn't complain either 
> which was logical.
> 
>     You could try with an older arm gdb to see if it works fine. If it does
>     and the latest version doesn't, then we may need to investigate it.
> 
> Ok, I'll compile and older gdb for arm and let you know.
> 
> 
>     I checked the code and the truncation message is a result of the
>     debugging stub sending less bytes than GDB expects for the register set.
> 
> The interesting thing is that the incompatible gdb for x86-64 was able 
> to read the message and the gdb for arm wasn't, but i'll try with an 
> older gdb for arm, but it's weird.

You might have gotten lucky. Or GDB wasn't too restrictive on the checks.

> 
>     Just to confirm, does your arm target have floating point registers?
> 
> Good point, I did assume it has, because in some cases armeabihf was 
> used to compile, but not in all cases, I'll check whether it is, 
> otherwise I'd have to recompile gdb without hf.
> I'll get the info and let you know, thanks for helping out!

Well, it seems GDB assumes floating point registers are there by 
default, so there wouldn't be a need to build GDB specifically for that.