Re: RFC: automatic dependency tracking for gdbserver

[Pedro Alves <palves@redhat.com>]

On 12/03/2012 09:17 PM, Tom Tromey wrote:
> This patch applies the same automatic dependency tracking logic that gdb
> uses to gdbserver.
> 
> The downside of this patch is that gdbserver dependencies will not be
> tracked unless the developer is using GNU make.  On the other hand,
> we've had this same code in gdb for several years without any
> complaints.
> 
> Comments?

Err, I love you? :-)  Seriously, many thanks for doing this.

> +amd64-linux-ipa.o : amd64-linux.c

Spurious space -----^

> diff --git a/gdb/gdbserver/aclocal.m4 b/gdb/gdbserver/aclocal.m4
> index 7b546b7..55bf0d8 100644
> --- a/gdb/gdbserver/aclocal.m4
> +++ b/gdb/gdbserver/aclocal.m4
> @@ -106,4 +106,6 @@ AC_DEFUN([_AM_SUBST_NOTMAKE])
>  # Public sister of _AM_SUBST_NOTMAKE.
>  AC_DEFUN([AM_SUBST_NOTMAKE], [_AM_SUBST_NOTMAKE($@)])
>  
> +m4_include([../../config/depstand.m4])
> +m4_include([../../config/lead-dot.m4])
>  m4_include([acinclude.m4])

I think I'd mildly prefer having these listed explicitly in acinclude.m4,
so you can do just autoreconf on the command line or aclocal, without worrying
about forgetting -I (IIRC, ACLOCAL_AMFLAGS is actually ignored, because we're
not using automake).  WDYT?  I actually have a patch for GDB that does the same
to gdb's aclocal.  See the second patch at:

  https://github.com/palves/gdb/commits/aclocal

but I never posted it.

I do wonder why does acinclude.m4 use sinclude instead of include.  I don't
see why you'd ever want those includes to silently fail.

-- 
Pedro Alves