This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFA 4/8] New port: TI C6x: Read loadmap from gdbserver
- From: Mark Kettenis <mark dot kettenis at xs4all dot nl>
- To: yao at codesourcery dot com
- Cc: vapier at gentoo dot org, gdb-patches at sourceware dot org
- Date: Mon, 8 Aug 2011 10:28:28 +0200 (CEST)
- Subject: Re: [RFA 4/8] New port: TI C6x: Read loadmap from gdbserver
- References: <4E263865.2070100@codesourcery.com> <CAJaTeTo6Yzqjq=L1=fXvgQmJb+UZNtsfRQtXyHY580YNTTZBFg@mail.gmail.com> <4E2CF0C6.2060504@codesourcery.com> <201108072030.02712.vapier@gentoo.org> <4E3F4E05.6010409@codesourcery.com>
> Date: Mon, 08 Aug 2011 10:46:29 +0800
> From: Yao Qi <yao@codesourcery.com>
>
> On 08/08/2011 08:30 AM, Mike Frysinger wrote:
> > On Monday, July 25, 2011 00:27:50 Yao Qi wrote:
> >> On 07/22/2011 10:16 AM, Mike Frysinger wrote:
> >>>> +#if defined __DSBT__
> >>>> +static int
> >>>
> >>> rather than being tied to the exec format that *gdbserver* is being
> >>> built as, shouldnt this be bound to the ptrace defines being available
> >>> ? how abut using "#ifdef PTRACE_GETDSBT" ?
> >>
> >> Yeah, that makes sense. Done.
> >
> > i think you missed a spot. one place uses __DSBT__ while another uses
> > PTRACE_GETDSBT.
>
> Yes, I thought it is not quite related to PTRACE stuff, so I didn't
> change that. I agree that we should use macros in a consistent way. I
> replace PTRACE_GETDSBT and __DSBT__ with PT_GETDSBT, because PT_GETDSBT
> is defined in sys/ptrace.h, but PTRACE_GETDSBT is defined in asm/ptrace.h.
Well, that leads to the following funny situation:
> +#if defined PT_GETDSBT
You check for PT_GETDSBT here...
> +static int
> +linux_read_loadmap (const char *annex, CORE_ADDR offset,
> + unsigned char *myaddr, unsigned int len)
> +{
> + int pid = lwpid_of (get_thread_lwp (current_inferior));
> + int addr = -1;
> + struct target_loadmap *data = NULL;
> + unsigned int actual_length, copy_length;
> +
> + if (strcmp (annex, "exec") == 0)
> + addr= (int) PTRACE_GETDSBT_EXEC;
> + else if (strcmp (annex, "interp") == 0)
> + addr = (int) PTRACE_GETDSBT_INTERP;
> + else
> + return -1;
> +
> + if (ptrace (PTRACE_GETDSBT, pid, addr, &data) != 0)
...but use PTRACE_GETDSBT here.
Now the kernel vs. libc headers issue has always been a contentious
one on Linux. But I think you should use the PTRACE_-prefixed names
in your code since those are the "official" Linux names, since Linux
was intended to be System V compatible. The PT_-prefixed names are
really only for compatibility with BSD (So I don't really understand
why people keep adding them for ptrace(2) requests that no BSD variant
ever had).