Re: FreeBSD/alpha kern.flp flood

From: Bruce Evans <bde_at_zeta.org.au>
Date: Wed, 2 Apr 2003 22:07:28 +1000 (EST)
On Wed, 2 Apr 2003, Ruslan Ermilov wrote:

> On Wed, Apr 02, 2003 at 03:24:59PM +1000, Bruce Evans wrote:
> > On Tue, 1 Apr 2003, David O'Brien wrote:
> >
> > > On Mon, Mar 31, 2003 at 06:03:48PM +0300, Ruslan Ermilov wrote:
> > > > Anyway, I've committed the fix to release/Makefile that strips
> > > > the .comment section out from the BOOTMFS kernel.  If you have
> > > > ...
> > > Why can't the bits be built with -Dlint to remove rcsid's.
> >
> > This would make little difference, since using __FBSDID() or rcsid[]
> > in kernel sources is a style bugs, and most of the sources are missing
> > this style bug.
> >
> I'd hardly call it a bug, since style(9) explicitly says C files
> should use __FBSDID().

Another bug.  __FBSDID() is is not normally used in the kernel.  E.g.,
rev.1.1 of almost every file in libkern is about removing (sccs) id
strings since their bloat is not wanted in the kernel.  It is true
that __FBSDID() and suitable stripping can handle this more configurably,
but no one ever cared enough to change n,000 $FreeBSD$'s to __FBSDID()'s
in the kernel (ugh) or provide infrastructure (#define lint doesn't count)
for stripping the ids.

> > Most of the savings from stripping commits is from removing verbose compiler
> > id "GCC: (GNU) 3.2.1 [FreeBSD] 20021119 (release)".
> >
> And I wondered why I get so huge savings where only few files
> put $FreeBSD$ into a .comment by way of __FBSDID().  Sure,
> ``objdump -s -j .comment kernel'' is full of these.
>
> David, can we get rid of the .comment section for the normal
> builds too, or at least not put these long GCC strings into?

They bloat programs and libraries too.  I once saved a MB or so in the
root partition by just by removing a couple of useless words in these
ids.  I like having these ids in object files and static libraries but
not duplicating them in executables and shared libraries.

Bruce
Received on Wed Apr 02 2003 - 02:07:37 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:02 UTC