From owner-svn-ports-head@FreeBSD.ORG Fri Aug 15 21:55:38 2014 Return-Path: Delivered-To: svn-ports-head@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [8.8.178.115]) (using TLSv1 with cipher ADH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id 5762AF1B for ; Fri, 15 Aug 2014 21:55:38 +0000 (UTC) Received: from freefall.freebsd.org (freefall.freebsd.org [IPv6:2001:1900:2254:206c::16:87]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id 2CA3020BA for ; Fri, 15 Aug 2014 21:55:38 +0000 (UTC) Received: from freefall.freebsd.org (localhost [127.0.0.1]) by freefall.freebsd.org (8.14.9/8.14.9) with ESMTP id s7FLtbKT058280 for ; Fri, 15 Aug 2014 21:55:37 GMT (envelope-from bdrewery@freefall.freebsd.org) Received: (from bdrewery@localhost) by freefall.freebsd.org (8.14.9/8.14.9/Submit) id s7FLtbsV058276 for svn-ports-head@freebsd.org; Fri, 15 Aug 2014 21:55:37 GMT (envelope-from bdrewery) Received: (qmail 266 invoked from network); 15 Aug 2014 16:55:35 -0500 Received: from unknown (HELO ?10.10.0.24?) (freebsd@shatow.net@10.10.0.24) by sweb.xzibition.com with ESMTPA; 15 Aug 2014 16:55:35 -0500 Message-ID: <53EE81CC.2080406@FreeBSD.org> Date: Fri, 15 Aug 2014 16:55:24 -0500 From: Bryan Drewery Organization: FreeBSD User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:31.0) Gecko/20100101 Thunderbird/31.0 MIME-Version: 1.0 To: Vsevolod Stakhov , Matthias Andree , Tijl Coosemans Subject: Re: svn commit: r364287 - head/ports-mgmt/pkg-devel References: <53e39939.55bc.4ca5432c@svn.freebsd.org> <20140807172841.58633e63@kalimero.tijl.coosemans.org> <53E3A468.5050603@FreeBSD.org> <53E3AC0C.5020904@gmx.de> <53E3AD09.2050000@FreeBSD.org> <53E3B3B5.9000104@gmx.de> <53E3B6D8.9080101@FreeBSD.org> In-Reply-To: <53E3B6D8.9080101@FreeBSD.org> OpenPGP: id=6E4697CF; url=http://www.shatow.net/bryan/bryan2.asc Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="bs2LnffeBW4FFJCBMH1eCSRXLAbraTQvU" Cc: svn-ports-head@freebsd.org, svn-ports-all@freebsd.org, baptiste Daroussin , tijl Coosemans , ports-committers@freebsd.org X-BeenThere: svn-ports-head@freebsd.org X-Mailman-Version: 2.1.18-1 Precedence: list List-Id: SVN commit messages for the ports tree for head List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 15 Aug 2014 21:55:38 -0000 This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --bs2LnffeBW4FFJCBMH1eCSRXLAbraTQvU Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: quoted-printable On 8/7/2014 12:26 PM, Vsevolod Stakhov wrote: > On 07/08/14 18:13, Matthias Andree wrote: >> Am 07.08.2014 um 18:44 schrieb Vsevolod Stakhov: >> >>> On 07/08/14 17:40, Matthias Andree wrote: >>>> Am 07.08.2014 um 18:08 schrieb Vsevolod Stakhov: >>>> >> [skipped] >> Please reconsider/revert that change, and instead snatch the embedded >> ELF SONAME from all libraries and use that for shared object (=3D dyna= mic >> library) requires/provides. >=20 > Ok, I'll spend some more of my time to explain the problem. >=20 > 1) We have the field called manifestdigest, that is intended to be > unique per package. > 2) We have packages with wrong provides (they could be generated by 1.2= > or some of 1.3 versions) > 3) Manifestdigest is not changed by shlib changes (there is a request > but it requires to rebuild all digests from the scratch). > 4) So for now, we have installed package with a wrong provide and a > remote package with the same manifestdigest. > 5) Remote package is not even considered by pkg as packages with the > same digests are considered equal. > 6) Pkg installs something from the remote repo that has the provide > needed (e.g. wine-devel-i386). >=20 > Hence, I *won't* revert this change until we fix other problems with > shared libraries and digests calculation. >=20 > I understand that it might be unclear, but currently shared library > dependencies are used merely to fix ports that lack explicit > dependencies. For example, if your port has something like >=20 > LIB_DEPENDS=3D libblah.so.2:${PORTSDIR}/some/port >=20 > then you would have *explicit* dependency and you won't be affected by > my change. It merely prevents pkg from installing unnecessary shit for = now. >=20 I am reverting this. I never should have released it. It has created a really bad situation. All packages now register themselves as providing "libblah.so" and requiring "libblah.so" without a version. This destroys all possibility of using versions in shlibs without recreating all packages and running 'pkg check -Ba' on systems. pkg query/pkg rquery %B/%b is now useless. To workaround this I am going to add a counter var to ports-mgmt/pkg/Makefile that poudriere can use to note what "version" packages are built from. If we bump it then poudriere will force rebuild all packages. As for 'pkg check -B' I am considering adding it in as an automatic post-pkg-upgrade task. I incorrectly assumed this change was only in the *solver* and not in the registered package. I should have reviewed/tested this. This change is not right. It is analogous to stripping package versions and using only names. I think we can all agree how crazy that would be. Looking over several ports I see no discrepancies between what a binary registers as requiring and what a shlib package advertises as provided. They match. Some packages have bogus shlibs provided, such as lang/ruby19 providing "openssl.so" (among many others) because (as you noted yourself) we look outside normal library dir search paths. However, this should not cause problems as packages won't be requiring openssl.so, they will require a versioned lib. If the solver needs to ignore shlibs (now that manifestdigest considers it) then lets do that. But we can't strip versions in the manifests, and ignoring the version for solving is not right either. Reviewing IRC logs it still seems to me that this all came about due to Wine. The only bit I can find otherwise is you mentioning a libxml2 discrepancy (which even mentions wine): > DBG(2)[4460]> require rule: (i386-wine-devel~emulators/i386-wine-devel(= r) | i386-wine~emulators/i386-wine(r) | !libxslt~textproc/libxslt(r)) > DBG(2)[5420]> libxml2.so.2 rule: (i386-wine-devel~emulators/i386-wine-d= evel(r) | i386-wine~emulators/i386-wine(r) | !libxslt~textproc/libxslt(r)= ) > DBG(3)[5420]> added shlib provide libxml2.so.2.9.1 for textproc/libxml2= > DBG(3)[5420]> added shlib deps for textproc/libxml2 on libxml2.so.2 Reinstalling textproc/libxml2 shows it is providing libxml2.so.2: # pkg query %b libxml2 libxml2.so.2 # readelf -d /usr/local/lib/libxml2.so.2.9.1|grep SONAME 0x000000000000000e (SONAME) Library soname: [libxml2.so.2] I'm not sure where libxml2.so.2.9.1 came from in your test but I cannot recreate it now. Running readelf -d against /usr/local/lib/*.so on my system shows all (except boost) have the simpler .so.X SONAMEs rather than .so.X.Y.Z. @tijl: Did something with SONAME change recently? --=20 Regards, Bryan Drewery --bs2LnffeBW4FFJCBMH1eCSRXLAbraTQvU Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (MingW32) iQEcBAEBAgAGBQJT7oHMAAoJEDXXcbtuRpfPek8IANV/9NV78ktiKmy5n5EE8oUw wESaw5MD+t5Ph2Ls2AdfO0LqDnOUrxZ2X+iP0aCe/zwioBRvbEi1VjJ7t61WZtdp lcUpJDd8h1lpXP4F8DVHvAUaBLcx4+wkkzJUYRJ7f7KY9wTK6s5LJZh5Nh3UsbIe +aRs7I1h8qMvFfPajrfRHm+7+AtalNyAn+xSoiZwZNk6alpdIfeZ50g6M3j32Kt2 3tl0VcmYb02otWN38TZ7kPMXe3wScgxEO+xqJ/yfMc3qT2SEDUiCfmXVoW/qZnGZ n2Cd/mASk4M+K07Q4Se9/sWsmdRZqTPd35qMQDW2sHr5QZav6k+zjpnoCKqc6vA= =jlgo -----END PGP SIGNATURE----- --bs2LnffeBW4FFJCBMH1eCSRXLAbraTQvU--