Date: Tue, 20 Dec 2005 15:58:22 +0000 From: Ceri Davies <ceri@submonkey.net> To: John Baldwin <jhb@freebsd.org> Cc: freebsd-hackers@freebsd.org Subject: Re: Mostly static binaries with crunchgen Message-ID: <20051220155821.GG68989@submonkey.net> In-Reply-To: <200512201029.28699.jhb@freebsd.org> References: <20051220114121.GA58620@submonkey.net> <200512201029.28699.jhb@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--6e7ZaeXHKrTJCxdu Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Tue, Dec 20, 2005 at 10:29:27AM -0500, John Baldwin wrote: > On Tuesday 20 December 2005 06:41 am, Ceri Davies wrote: > > Adrian Steinmann's talk at EuroBSDcon regarding a single user SSH daemon > > for rescue purposes highlighted an interesting point regarding some > > binaries. The GEOM userland binaries such as gmirror, gstripe, etc. use > > dlopen() to load classes from /lib/geom and therefore cannot be > > statically linked and, by extension, cannot be crunched with crunchgen. > > > > Adrian mentioned that it would be useful if crunchgen(1) supported > > "mostly static" binaries; i.e., a libs_so extension to crunchgen which > > would allow these binaries to be crunched, simply requiring then that > > rtld and the libraries be made available on the memory disk. This > > allows those of us who use GEOM classes to make a small rescue disk. > > > > I started to add this to crunchgen on the way home, and have worked with > > Adrian to finish it off. The patch is attached. It simply adds a > > "libs_so" keyword which specifies libraries that will be linked > > statically; all current config files continue to produce the same code > > as they did before. > > > > I'd like to commit this with a 6 week MFC period or so, but my mentor is > > currently busy. Could someone else please take this up? >=20 > I don't think you should change TORTIOUS to TORTUOUS in the license. Rea= ding=20 > license disclaimers may indeed be tortuous, but tortious is an actual leg= al=20 > term, not a misspelling. It comes from the root word 'tort' which is a l= egal=20 > word for 'sue' (basically). Ah, I had parsed that as a whitespace change. Will revert :) > The other concern is does this force the entire crunch to require a worki= ng=20 > rtld now? If so, that would mean that this wouldn't be appropriate for= =20 > something such as /rescue. If there were a way to statically link rtld i= nto=20 > the crunch itself that would probably be ideal, but I'm not sure that is= =20 > possible. No, just the dynamic bits require rtld. Ceri --=20 Only two things are infinite, the universe and human stupidity, and I'm not sure about the former. -- Einstein (attrib.) --6e7ZaeXHKrTJCxdu Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2 (FreeBSD) iD8DBQFDqCodocfcwTS3JF8RAi9UAJ92xNID3W5kwjB7t+xy1i5t5OUZgACZAZFd M4US9ShP+T2RKKmm5sMAvp4= =KrrR -----END PGP SIGNATURE----- --6e7ZaeXHKrTJCxdu--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051220155821.GG68989>