Re: [PATCH 2/2] Implement pid_to_exec_file for Windows in gdbserver

[Simon Marchi <simark@simark.ca>]

> diff --git a/gdb/nat/windows-nat.h b/gdb/nat/windows-nat.h
> index 29fd0a3a69b..4244426f5eb 100644
> --- a/gdb/nat/windows-nat.h
> +++ b/gdb/nat/windows-nat.h
> @@ -249,6 +249,8 @@ struct windows_process_info
>  
>    gdb::optional<pending_stop> fetch_pending_stop (bool debug_events);
>  
> +  const char *pid_to_exec_file (int);

It seems a bit strange to pass a pid to this method.  Doesn't a
windows_process_info object know its own pid?  And so passing the pid as
a parameter would be redundant?

Right now gdb/windows-nat.c supports a single process only, so it must
be that the pid that is requested as a parameter must match the (only)
process' pid.  So maybe we can have a gdb_assert in
windows_nat_target::pid_to_exec_file to verify that (and the same
GDBserver-side)?

The day the windows-nat target becomes multi-process, then I guess we'll
use the passed-in pid to look up a specific windows_process_info by pid
in whatever data structure will contain all the windows_process_infos.
And then we can call windows_process_info::pid_to_exec_file without
passing any pid.

Note that this is all speculation, I haven't tried anything.  I'm just
assuming that windows_process_info::id is the same as "pid".

Otherwise, LGTM.

Simon