https://bugzilla.redhat.com/show_bug.cgi?id=837004
--- Comment #5 from Paul Howarth <paul(a)city-fan.org> ---
(In reply to comment #1)
Building on an installed system (to check for possible conflicts),
not a
buildroot I got this:
RPM build errors:
Installed (but unpackaged) file(s) found:
/usr/share/emacs/site-lisp/pari/pari-completion.el
/usr/share/emacs/site-lisp/pari/pari-completion.elc
/usr/share/emacs/site-lisp/pari/pari-conf.el
/usr/share/emacs/site-lisp/pari/pari-conf.elc
/usr/share/emacs/site-lisp/pari/pari-fontification.el
/usr/share/emacs/site-lisp/pari/pari-fontification.elc
/usr/share/emacs/site-lisp/pari/pari-help.el
/usr/share/emacs/site-lisp/pari/pari-messages.el
/usr/share/emacs/site-lisp/pari/pari-messages.elc
/usr/share/emacs/site-lisp/pari/pari.el
/usr/share/emacs/site-lisp/pari/pari.elc
/usr/share/emacs/site-lisp/pari/pariemacs.txt
/usr/share/emacs/site-lisp/pari/sli-tools.el
/usr/share/emacs/site-lisp/pari/sli-tools.elc
I suggest something like:
$ diff -u SPECS/libpari23.spec 837004/libpari23.spec
--- SPECS/libpari23.spec 2012-07-02 09:09:50.000000000 -0400
+++ 837004/libpari23.spec 2012-07-02 11:31:53.768642232 -0400
@@ -118,6 +118,9 @@
# Placate rpmlint regarding binary and library permissions
%{_fixperms} %{buildroot}{%{_bindir},%{_libdir}}
+# Cleanup emacs in alternate package
rm -fr %{buildroot}%{_datadir}/emacs
+
I've added something like this in -2.
After the above, you will need to update perl-Math-Pari.spec, to
avoid
$ rpmbuild -ba SPECS/perl-Math-Pari.spec
error: Failed build dependencies:
pari-devel = 2.3.5 is needed by
perl-Math-Pari-2.010806-13.fc18.x86_64
somewhat like:
$ diff -u SPECS/perl-Math-Pari.spec{~,}
--- SPECS/perl-Math-Pari.spec~ 2012-07-02 11:35:37.129650783 -0400
+++ SPECS/perl-Math-Pari.spec 2012-07-02 11:36:57.368653861 -0400
@@ -30,7 +30,7 @@
Requires: perl(:MODULE_COMPAT_%(eval "`perl -V:version`"; echo
$version))
# Enforce dependency against same version of pari that we're built for
-BuildRequires: pari-devel = %{pari_version}
+BuildRequires: libpari23-devel
Requires: pari%{?_isa} = %{pari_version}
# Don't "provide" private Perl libs or the redundant unversioned
perl(Math::Pari)
after the above, I only did need:
$ diff -u SPECS/perl-Math-Pari.spec{.orig,}
--- SPECS/perl-Math-Pari.spec.orig 2012-07-02 11:36:57.368653861 -0400
+++ SPECS/perl-Math-Pari.spec 2012-07-02 11:40:12.704661293 -0400
@@ -62,8 +62,8 @@
%build
perl Makefile.PL \
INSTALLDIRS=vendor \
- OPTIMIZE="-I%{_includedir}/pari %{optflags}" \
- parilib="-lpari"
+ OPTIMIZE="-I%{_includedir}/libpari23/pari %{optflags}" \
+ parilib="-lpari23"
make %{?_smp_mflags}
I've got a re-worked perl-Math-Pari package here, with quite a few changes for
moving to libpari23:
http://www.city-fan.org/~paul/extras/pari/perl-Math-Pari-2.010806-14.fc18...
(In reply to comment #2)
[!]: MUST Package has no %clean section with rm -rf %{buildroot} (or
$RPM_BUILD_ROOT)
Note: Clean is needed only if supporting EPEL
Fixed in -2
[x]: MUST Package does not run rm -rf %{buildroot} (or
$RPM_BUILD_ROOT) at
the
beginning of %install.
Fixed in -2
[!]: MUST Package requires pkgconfig, if .pc files are present.
(EPEL5)
Note: Only applicable for EL-5
Didn't remove the explicit requirement here - I think it's harmless anyway.
[!]: MUST Rpmlint output is silent.
rpmlint libpari23-2.3.5-1.fc18.i686.rpm
libpari23.i686: W: spelling-error %description -l en_US factorizations ->
factorization, factorization s, categorizations
False positive.
libpari23.i686: E: shlib-with-non-pic-code
/usr/lib/libpari-gmp.so.2.3.5
Fixed in -2
libpari23.i686: W: shared-lib-calls-exit
/usr/lib/libpari-gmp.so.2.3.5
exit(a)GLIBC_2.0
That's one for upstream
libpari23.i686: E: incorrect-fsf-address
/usr/share/doc/libpari23-2.3.5/COPYING
Not allowed to change that for legal reasons. Already fixed in upstream VCS.
rpmlint libpari23-debuginfo-2.3.5-1.fc18.i686.rpm
libpari23-debuginfo.i686: E: incorrect-fsf-address
/usr/src/debug/pari-2.3.5/src/graph/plotport.c
These are all fixed in -2
rpmlint libpari23-2.3.5-1.fc18.src.rpm
libpari23.src: W: spelling-error %description -l en_US factorizations ->
factorization, factorization s, categorizations
False positive again
libpari23.src:93: E: hardcoded-library-path in %{_prefix}/lib/pari
The place where this happens is where we remove files that upstream has put
there.
rpmlint libpari23-devel-2.3.5-1.fc18.i686.rpm
libpari23-devel.i686: W: no-documentation
There but it's not marked as %doc as it's needed by other packages so we
don't
want it omitted when installed using --nodocs
libpari23-devel.i686: E: incorrect-fsf-address
/usr/share/libpari23/src/headers/pariport.h
These are all fixed in -2
libpari23-devel.i686: E: zero-length
/usr/share/libpari23/src/test/32/qfbsolve
libpari23-devel.i686: E: zero-length /usr/share/libpari23/src/test/32/rfrac
I think these are intentionally zero length
libpari23-devel.i686: E: zero-length
/usr/share/libpari23/doc/libpari.toc
libpari23-devel.i686: E: zero-length /usr/share/libpari23/doc/refcard.idx
libpari23-devel.i686: E: zero-length /usr/share/libpari23/doc/tutorial.idx
libpari23-devel.i686: E: zero-length /usr/share/libpari23/doc/develop.aux
libpari23-devel.i686: E: zero-length /usr/share/libpari23/doc/users.toc
libpari23-devel.i686: E: zero-length /usr/share/libpari23/doc/develop.toc
libpari23-devel.i686: E: zero-length /usr/share/libpari23/doc/refcard.aux
This is all fall-out from the tex documentation build; not sure if omitting
them would cause problems for downstream users (Math::Pari builds its own POD
documentation from the tex files).
libpari23-devel.i686: E: wrong-script-interpreter
/usr/share/libpari23/doc/gphelp.in @perl@
This is intended to be substituted by downstream users.
[-]: SHOULD Latest version is packaged.
This is intentionally not the latest version, for compatibility reasons.
[!]: SHOULD SourceX / PatchY prefixed with %{name}.
Note: Source0:
http://pari.math.u-bordeaux.fr/pub/pari/unix/OLD/pari-%{version}.tar.gz
(pari-%{version}.tar.gz) Patch1: libpari23-optflags.patch
(libpari23-optflags.patch) Patch2: Math-Pari-2.01080604-extra-stack-for-
test.patch (Math-Pari-2.01080604-extra-stack-for-test.patch)
Patch2 has origins in the perl-Math-Pari package, hence its prefix.
> Issues:
[!]: MUST Package has no %clean section with rm -rf %{buildroot} (or
$RPM_BUILD_ROOT)
Note: Clean is needed only if supporting EPEL
> See:
http://fedoraproject.org/wiki/Packaging/Guidelines#.25clean
> [!]: MUST Package does not run rm -rf %{buildroot} (or $RPM_BUILD_ROOT) at
> the
> beginning of %install.
> Note: rm -rf is only needed if supporting EPEL5
> See: None
[!]: MUST Package requires pkgconfig, if .pc files are present.
(EPEL5)
Note: Only applicable for EL-5
> See:
http://fedoraproject.org/wiki/EPEL/GuidelinesAndPolicies#EL5
> [!]: MUST Rpmlint output is silent.
I think these are addressed in -2 as described above.
Spec URL:
http://subversion.city-fan.org/repos/cfo-repo/libpari23/branches/fedora/l...
SRPM URL:
http://www.city-fan.org/~paul/extras/pari/libpari23-2.3.5-2.fc18.src.rpm
(In reply to comment #4)
Ops, doing my homework, updating flags and assigning to myself.
Waiting for some information (actually, only I am only concerned about
%clean) to change from review-? to review-+
But I think there may be issues with the libpari23.pc file, e.g.
includes are under %{_includedir}/libpari23/pari and I think most code using
pari does not include "pari/name.h" but "name.h"
This applies for perl-Math-Pari for instance. However, upstream's API
documentation for the current release says to use <pari/pari.h>, which is why I
went the way I did. Looking back at older documentation, it seems this has
changed since 2.3.0 where the recommendation was just <pari.h> so I'm open to
changing it but I don't think it's a big deal.
BTW, yet another issue, also in pari.spec, is that pari.cfg should
not be
marked %doc. But not a big issue, as I believe no code linking to pari cares
about it.
That's what I thought too, which is why I just included in %{_docdir} rather
than under %{_sysconfdir}.
--
You are receiving this mail because:
You are on the CC list for the bug.