[Lldb-commits] [lldb] [lldb][Windows] Fix "Invalid register name" for eax on x86_64 attach (PR #203498)

[Jason Molenda via lldb-commits]

jasonmolenda wrote:

I'm still a little confused by why this is needed.  `BuildDynamicRegisterInfo` should call qProcessInfo itself; I think you're working around some bug in the way `BuildDynamicRegisterInfo` determines the arch by moving it to the end of this method.  I'm sure it works, but I'm genuinely curious why this block in `BuildDynamicRegisterInfo,`

```
   471    const ArchSpec &target_arch = GetTarget().GetArchitecture();
   472    const ArchSpec &remote_host_arch = m_gdb_comm.GetHostArchitecture();
   473    const ArchSpec &remote_process_arch = m_gdb_comm.GetProcessArchitecture();
   474  
   475    // Use the process' architecture instead of the host arch, if available
   476    ArchSpec arch_to_use;
   477    if (remote_process_arch.IsValid())
   478      arch_to_use = remote_process_arch;
   479    else
   480      arch_to_use = remote_host_arch;
   481  
   482    if (!arch_to_use.IsValid())
   483      arch_to_use = target_arch;
   484  
   485    llvm::Error register_info_err = GetGDBServerRegisterInfo(arch_to_use);
```

doesn't work.  On my macOS system, `m_gdb_comm.GetProcessArchitecture()` issues a `qProcessInfo` and gets a correct ArchSpec from that.  That must be failing on the Windows system, and we fall back to using the ArchSpec from `qHostInfo` (`m_gdb_comm.GetHostArchitecture()`) which is also incorrect?  And we don't have a Target arch yet, which is fine, we haven't loaded a binary yet.

https://github.com/llvm/llvm-project/pull/203498