Re: [PROPOSAL] changing policy on compiling with -g .. a better way
On Tue, Aug 31, 1999 at 10:44:49AM -0500, Manoj Srivastava wrote:
> Hi,
> >>"Ben" == Ben Collins <bcollins@debian.org> writes:
>
> Ben> As a buildd admin, I want to congratulate the original policy on
> Ben> all the wasted cpu cycles it has cost my system by forcing
> Ben> packages to compile with -g even though those same binaries will
> Ben> be stripped later of this costly debugging information.
>
> Starting a new proposal with sarcasm and belittling the
> original policy process is not very helpful
Lighten up, this is all supposed to be fun.
> Ben> -------- The package can by default build without -g if it also
> Ben> provides a mechanism to easily be rebuilt with debugging
> Ben> information. This can be done by providing a "build-debug" make
> Ben> target, or allowing the user to specify "BUILD_DEBUG=yes" in the
> Ben> environment while compiling that package.
> Ben> --------
>
> I think we should specify one, or the other, or both, so that
> people do not have to grok rules files to determine which m,ethod to
> use. The, one can also set build daemons to build a debuggable
> distribution if one so desires.
Already talked about and settled on.
> However, have you looked at the cost of this proposal? This
> entails that one massage upstream Makefiles (or several Makefiles) to
> take not of an environment variable to add debugging flags. That is
> more difficult than a static, one time edit of the Makefiles involved
> to add the -g and the strip commands.
>
> Please note that you may have to take additional action to
> pass these variables to a sub-make. Except by explicit request,
> `make' exports a variable only if it is either defined in the
> environment initially or set on the command line, and if its name
> consists only of letters, numbers, and underscores.
* This is not a proposal to force poeple to change what they already have
but more of a way to let them do it a different way. I'm not proposing
that every package should do this, that would be insane.
So it's harder to do this than to force everyone to use "-g -O2" ? I don't
see that argument, but my statement above makes it irrelevant anyway.
> Could people on this list try it on their packages and see how
> hard this would be? That would give us a cross section of all
> packages, hopefully.
>
> I would rather we have some numbers before we agre to do this.
It worked on all of my packages. Granted they all use autoconf and they
all honor CFLAGS, but most packages do aswell. THe ones that don't, can
stay just the way they are.
Ben
Reply to: