Re: SPECS (rpm-4_5): rpm.spec - deem. need different sources for db due md5
On Sep 3, 2008, at 2:57 PM, Jakub Bogusz wrote: > On Wed, Sep 03, 2008 at 07:17:56PM +0300, Elan Ruusamäe wrote: >> On Wednesday 03 September 2008 18:15:30 Jakub Bogusz wrote: >>> Serveral issues now: >>> - It makes rpm sources weight almost 30MB(!) (compared to 7MB from >>> upstream, which IIRC include yet one more copy of db sources) >> >> however i have some other questions: >> >> would it be ok to enable --uniquename in db-static build? (in >> db*.spec) > > Would be acceptable in -static, but it would require some function > names > mangling in header(?). > Don't try it in shared (as long as it's named just libdbX.Y). > (glen asked me on #rpm to explain --with-uniqname usage) If compiled using --with-unquename, there are symbol redefine's added in db.h, basically adding --with-uniqname=arg suffix. So if --with-uniqname is used for "system" static libraries, one would also have a custom db.h that would need to be added somewhere. >> how do you tell automake to link with libdb.a, but not linking >> everything >> static (ie other libs would be dynamic linking) > > Either: > - specify full pathname ( $(libdir)/libdb.a ) > - as libtool will probably break "-Wl,-Bstatic -ldb -Wl,-Bdynamic", > you > can try something like "-Wl,-Bstatic,-ldb,-Bdynamic" or > "-Wl,-Bstatic -Wl,-ldb -Wl,-Bdynamic" > Static linkage is likely a viable option. However, note that all the db_utils and every rpm excutable that does -lrpmdb (there are more than rpm & rpmbuild that do -lrpmdb) will include a copy of Berkeley DB if statically linked. OTOH, the original bug report https://bugs.launchpad.net/pld-linux/ac/+bug/247567 appears to be from identical symbols from 2 versions of Berkeley DB, one from rpm + "system" db, the other from nss_db with internal db. Does nss_db use --with-uniqename? Or is there some symbol that isn't handled correctly if --with- uniqename that collides with rpm + "system" db? What is the poldek segfault backtrace? hth 73 de Jeff Does nss > > -- > Jakub Boguszhttp://qboosh.pl/ > ___ > pld-devel-en mailing list > pld-devel-en@lists.pld-linux.org > http://lists.pld-linux.org/mailman/listinfo/pld-devel-en ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: SPECS (rpm-4_5): rpm.spec - deem. need different sources for db due md5
On Sep 3, 2008, at 2:51 PM, Jakub Bogusz wrote: > > > As for db maintenance - I though about db changes themselves, not > clashes. > There are no changes to internal db in rpm. The possible exception is "robust mutex" handling, but one can live without that functionality. > Currently _this_ bug is worked around by internal db used in > nss_db, but > it seems more general problem with glibc importing too much from nss > module (think about different versions of mysql or ldap libs). > I'm not sure if it still exists with current glibc, IIRC I've seen > some > comments on libc-alpha or so. > Note that nss has its own bloat through internal db. So two faults == rpm error as always. > > BTW, just found funny rpm "bug" (originated from patch about 9 years > ago, when rpm didn't accept specs with non-ascii characters): > https://bugzilla.redhat.com/show_bug.cgi?id=226377#c14 > ;-) Yes hysterical. What is less hysterical is this bug: http://rpm5.org/cvs/tktview?tn=38,5 No encoding can be specified for tags in a *.spec file. And so stoopid checks like what the comment in bz#226377 is attached to are all that can be attempted 9+ years later. Sad but true. The underlying __DESIGN__ problem in rpm is that RPM_I18NSTRING_TYPE, an associative array using locale, to return a string with no obviious or reliable means to handle encoding transforms. But I digress ... 73 de Jeff > > -- > Jakub Boguszhttp://qboosh.pl/ > ___ > pld-devel-en mailing list > pld-devel-en@lists.pld-linux.org > http://lists.pld-linux.org/mailman/listinfo/pld-devel-en ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: SPECS (rpm-4_5): rpm.spec - deem. need different sources for db due md5
On Wed, Sep 03, 2008 at 07:17:56PM +0300, Elan Ruusamäe wrote: > On Wednesday 03 September 2008 18:15:30 Jakub Bogusz wrote: > > Serveral issues now: > > - It makes rpm sources weight almost 30MB(!) (compared to 7MB from > > upstream, which IIRC include yet one more copy of db sources) > > however i have some other questions: > > would it be ok to enable --uniquename in db-static build? (in db*.spec) Would be acceptable in -static, but it would require some function names mangling in header(?). Don't try it in shared (as long as it's named just libdbX.Y). > how do you tell automake to link with libdb.a, but not linking everything > static (ie other libs would be dynamic linking) Either: - specify full pathname ( $(libdir)/libdb.a ) - as libtool will probably break "-Wl,-Bstatic -ldb -Wl,-Bdynamic", you can try something like "-Wl,-Bstatic,-ldb,-Bdynamic" or "-Wl,-Bstatic -Wl,-ldb -Wl,-Bdynamic" -- Jakub Boguszhttp://qboosh.pl/ ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: SPECS (rpm-4_5): rpm.spec - deem. need different sources for db due md5
On Wed, Sep 03, 2008 at 07:15:16PM +0300, Elan Ruusamäe wrote:
> On Wednesday 03 September 2008 18:15:30 Jakub Bogusz wrote:
> > On Wed, Sep 03, 2008 at 04:44:34PM +0200, glen wrote:
> > > Author: glen Date: Wed Sep 3 14:44:34 2008 GMT
> > > Module: SPECS Tag: rpm-4_5
> > > Log message:
> > > - deem. need different sources for db due md5
> > >
> > > -Source19:
> > > http://download.oracle.com/berkeley-db/db-%{reqdb_ver}.tar.gz
> > > +Source19:http://download.oracle.com/berkeley-db/db-4.5.20.tar.gz
> > > # Source19-md5: b0f1c08cb8e9d37fb47e7ed3312d
> > > +Source20:http://download.oracle.com/berkeley-db/db-4.7.25.tar.gz
> > > +# Source20-md5: ec2b87e833779681a0c3a814aa71359e
> > > %patchset_source -f
> > > http://www.oracle.com/technology/products/berkeley-db/db/update/%{reqdb_v
> > >er}/patch.%{reqdb_ver}.%g 1 %{reqdb_patch}
> > > Patch1067:%{name}-disable-features.patch
> > > Patch1070: %{name}-rpmrc-ac.patch
> > > @@ -656,8 +658,12 @@
> > > %prep
> > > %setup -q
> > > %if %{with internal_db}
> > > +%if "%{pld_release}" == "th"
> > > +%{__tar} -zxf %{SOURCE20} -C db3 --strip-components=1
> > > +%else
> > > %{__tar} -zxf %{SOURCE19} -C db3 --strip-components=1
> > > %endif
> > > +%endif
> >
> > Uh oh. Haven't "internal db" madness reached its maximum yet?
> >
> > Serveral issues now:
> > - It makes rpm sources weight almost 30MB(!) (compared to 7MB from
> > upstream, which IIRC include yet one more copy of db sources)
> see below
>
> > - files (patches) list is conditional, not strictly defined by spec
> > alone
> isn't this against our policy that sources/patches shouldn't be conditional,
> or what do you mean?
That was the point, patch list shouldn't be conditional and %patchset gave
conditional results.
As for Ac/Th db sources, cost of including 10+MB of sources to avoid
spec branch is too big.
> > - using different version of db than system one requires to provide
> > db utils (db_recover, db_dump etc.) in that particular version as well
>
> i haven't had successful build with internal db, so no %files update yet...
>
> > What are profits?
> > None if version of internal db ever changes. This would mean need to
> > maintain some old db version even if Oracle stops.
>
> how do you suggest solving this bug?
> https://bugs.launchpad.net/pld-linux/ac/+bug/247567
As for db maintenance - I though about db changes themselves, not
clashes.
Currently _this_ bug is worked around by internal db used in nss_db, but
it seems more general problem with glibc importing too much from nss
module (think about different versions of mysql or ldap libs).
I'm not sure if it still exists with current glibc, IIRC I've seen some
comments on libc-alpha or so.
BTW, just found funny rpm "bug" (originated from patch about 9 years
ago, when rpm didn't accept specs with non-ascii characters):
https://bugzilla.redhat.com/show_bug.cgi?id=226377#c14
--
Jakub Boguszhttp://qboosh.pl/
___
pld-devel-en mailing list
pld-devel-en@lists.pld-linux.org
http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: SPECS (rpm-4_5): rpm.spec - deem. need different sources for db due md5
On Wednesday 03 September 2008 18:15:30 Jakub Bogusz wrote: > Serveral issues now: > - It makes rpm sources weight almost 30MB(!) (compared to 7MB from > upstream, which IIRC include yet one more copy of db sources) however i have some other questions: would it be ok to enable --uniquename in db-static build? (in db*.spec) how do you tell automake to link with libdb.a, but not linking everything static (ie other libs would be dynamic linking) -- glen ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: SPECS (rpm-4_5): rpm.spec - deem. need different sources for db due md5
On Wednesday 03 September 2008 18:15:30 Jakub Bogusz wrote:
> On Wed, Sep 03, 2008 at 04:44:34PM +0200, glen wrote:
> > Author: glen Date: Wed Sep 3 14:44:34 2008 GMT
> > Module: SPECS Tag: rpm-4_5
> > Log message:
> > - deem. need different sources for db due md5
> >
> > -Source19: http://download.oracle.com/berkeley-db/db-%{reqdb_ver}.tar.gz
> > +Source19: http://download.oracle.com/berkeley-db/db-4.5.20.tar.gz
> > # Source19-md5:b0f1c08cb8e9d37fb47e7ed3312d
> > +Source20: http://download.oracle.com/berkeley-db/db-4.7.25.tar.gz
> > +# Source20-md5:ec2b87e833779681a0c3a814aa71359e
> > %patchset_source -f
> > http://www.oracle.com/technology/products/berkeley-db/db/update/%{reqdb_v
> >er}/patch.%{reqdb_ver}.%g 1 %{reqdb_patch}
> > Patch1067: %{name}-disable-features.patch
> > Patch1070: %{name}-rpmrc-ac.patch
> > @@ -656,8 +658,12 @@
> > %prep
> > %setup -q
> > %if %{with internal_db}
> > +%if "%{pld_release}" == "th"
> > +%{__tar} -zxf %{SOURCE20} -C db3 --strip-components=1
> > +%else
> > %{__tar} -zxf %{SOURCE19} -C db3 --strip-components=1
> > %endif
> > +%endif
>
> Uh oh. Haven't "internal db" madness reached its maximum yet?
>
> Serveral issues now:
> - It makes rpm sources weight almost 30MB(!) (compared to 7MB from
> upstream, which IIRC include yet one more copy of db sources)
see below
> - files (patches) list is conditional, not strictly defined by spec
> alone
isn't this against our policy that sources/patches shouldn't be conditional,
or what do you mean?
> - using different version of db than system one requires to provide
> db utils (db_recover, db_dump etc.) in that particular version as well
i haven't had successful build with internal db, so no %files update yet...
> What are profits?
> None if version of internal db ever changes. This would mean need to
> maintain some old db version even if Oracle stops.
how do you suggest solving this bug?
https://bugs.launchpad.net/pld-linux/ac/+bug/247567
--
glen
___
pld-devel-en mailing list
pld-devel-en@lists.pld-linux.org
http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: SPECS (rpm-4_5): rpm.spec - deem. need different sources for db due md5
On Sep 3, 2008, at 11:15 AM, Jakub Bogusz wrote: > > What are profits? The profit is avoiding Berekely DB symbol clash by using --with- uniquename for rpm with internal db. Statically linking rpm will have larger bloatiness cost than what you mention. Whether you want that "profit" is a whole different question I'll leave to glen. 73 de Jeff ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: SPECS (rpm-4_5): rpm.spec - deem. need different sources for db due md5
On Wed, Sep 03, 2008 at 04:44:34PM +0200, glen wrote:
> Author: glen Date: Wed Sep 3 14:44:34 2008 GMT
> Module: SPECS Tag: rpm-4_5
> Log message:
> - deem. need different sources for db due md5
> -Source19:http://download.oracle.com/berkeley-db/db-%{reqdb_ver}.tar.gz
> +Source19:http://download.oracle.com/berkeley-db/db-4.5.20.tar.gz
> # Source19-md5: b0f1c08cb8e9d37fb47e7ed3312d
> +Source20:http://download.oracle.com/berkeley-db/db-4.7.25.tar.gz
> +# Source20-md5: ec2b87e833779681a0c3a814aa71359e
> %patchset_source -f
> http://www.oracle.com/technology/products/berkeley-db/db/update/%{reqdb_ver}/patch.%{reqdb_ver}.%g
> 1 %{reqdb_patch}
> Patch1067: %{name}-disable-features.patch
> Patch1070: %{name}-rpmrc-ac.patch
> @@ -656,8 +658,12 @@
> %prep
> %setup -q
> %if %{with internal_db}
> +%if "%{pld_release}" == "th"
> +%{__tar} -zxf %{SOURCE20} -C db3 --strip-components=1
> +%else
> %{__tar} -zxf %{SOURCE19} -C db3 --strip-components=1
> %endif
> +%endif
Uh oh. Haven't "internal db" madness reached its maximum yet?
Serveral issues now:
- It makes rpm sources weight almost 30MB(!) (compared to 7MB from upstream,
which IIRC include yet one more copy of db sources)
- files (patches) list is conditional, not strictly defined by spec
alone
- using different version of db than system one requires to provide
db utils (db_recover, db_dump etc.) in that particular version as well
What are profits?
None if version of internal db ever changes. This would mean need to
maintain some old db version even if Oracle stops.
--
Jakub Boguszhttp://qboosh.pl/
___
pld-devel-en mailing list
pld-devel-en@lists.pld-linux.org
http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: PLD Live 2.0 beta3
Dnia środa 03 wrzesień 2008, Michael Shigorin napisał: > The ISO is available here: > > ftp://ep09.pld-linux.org/people/qwiat/ > > Quite speedy Also here: ftp://pld-mirror.domainmaker.pl/pld-live/ -- Pozdrawiam, Best regards, Mit freundlichen Grüßen, Wojciech "Wojtosz" Błaszkowski ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
Re: PLD Live 2.0 beta3
On Sun, Aug 31, 2008 at 06:17:39PM +0200, Patryk Zawadzki wrote: > qwiat did a hell of a good job resurrecting the PLD LiveCD > idea. The ISO is available here: > ftp://ep09.pld-linux.org/people/qwiat/ Quite speedy :) -- WBR, Michael Shigorin <[EMAIL PROTECTED]> -- Linux.Kiev http://www.linux.kiev.ua/ ___ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en
