Re: enable/disable-debug options to libstdc++-v3

[Ranjit Mathew <rmathew at hotmail dot com>]

Phil Edwards wrote:
> On Mon, Oct 21, 2002 at 04:29:00PM +0530, Ranjit Mathew wrote:
>
> > Now I did read in the configure.in comments that
> > these flags should not be specified for production
> > builds, but the problem is that I need to specify
> > --disable-debug for other parts of GCC and the top-level
> > configure blindly passes all flags to all components.
> >
> > Can these flags not be changed to "-g0" when
> > "--enable-debug=no"?
> >
> > Does it really make sense to say that "--disable-debug"
> > would mean that debug information would be put in
> > anyways?
>
> The point of --enable-debug, when it was first written into libstdc++,
> was to add /extra/ debugging information, mostly for our own use.
>
> Are you certain that the increased size is due to debug information?
> If that is the case, then
>
>     1)  you can experiment with running 'strip' on the library, with
>         various options, and
>     2)  debug information is not loaded into memory at runtime; it takes
>         up extra space on disk, but the memory footprint will not change.
>
> Try running 'strip' with different options, and if the size changes
> significantly, then it's debug information, and if the size is the same,
> then it's something else.
Yes, I'm positive that it is the debug information - on MinGW
at least, the generated libstdc++-v3 is 14MB(!) with debug
info and just 800K without - on my personal GCC build environment,
I apply the following patch before building:
-------------------------- 8< ------------------------------
--- acinclude.m4        Tue Oct 22 09:43:14 2002
+++ acinclude.m4        Tue Oct 22 09:43:41 2002
@@ -1029,5 +1029,5 @@
         ;;
     no)
-        DEBUG_FLAGS='-g'
+        DEBUG_FLAGS='-g0'
         ;;
 esac
-------------------------- 8< ------------------------------

In fact, Danny Smith, the primary MinGW GCC/binutils maintainer
remarked in a post to the mingw-users list (where I'd raised
this issue) that he himself just uses "strip -g" on the library
before packaging it for distribution.

The difference in size might not impact run time performance
but it takes quite a bit of unnecessary disk space and has just
that little bit of impact at link time (IMHO).

Besides, as I remarked earlier, it is rather weird to say
that "--disable-debug" would put in a lot of debug
info anyways.

I haven't had the courage to see what monster is created
by the "--enable-debug" option. ;-)

Just my 2p.

Sincerely Yours,
Ranjit.

--
Ranjit Mathew          Email: rmathew AT hotmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/