rpm-4.6 and brp-python-bytecompile?

[Carl D. Roth]

I noticed this wierdness with rpm-4.6 (f10) and rpm-4.7 (f11)...  When I 
build packages including Python scripts, sometimes the .pyc and .pyo 
files are included in the package manifest, and sometimes, they are not.

The script that does the .pyc and .pyo generation is 'brp-python-
bytecompile', part of the '%__os_install_post' macro.  Here are the 
various (inconsistent) results I get on different systems:

F9:

$ rpm --eval '%__os_install_post'

    /usr/lib/rpm/redhat/brp-compress
    /usr/lib/rpm/redhat/brp-strip /usr/bin/strip
    /usr/lib/rpm/redhat/brp-strip-static-archive /usr/bin/strip
    /usr/lib/rpm/redhat/brp-strip-comment-note /usr/bin/strip /usr/bin/
objdump
    /usr/lib/rpm/brp-python-bytecompile
    /usr/lib/rpm/redhat/brp-python-hardlink
    /usr/lib/rpm/redhat/brp-java-repack-jars

F10 (x86_64):

$ rpm --eval '%__os_install_post'

    /usr/lib/rpm/redhat/brp-compress 
    /usr/lib/rpm/redhat/brp-strip /usr/bin/strip 
    /usr/lib/rpm/redhat/brp-strip-static-archive /usr/bin/strip 
    /usr/lib/rpm/redhat/brp-strip-comment-note /usr/bin/strip /usr/bin/
objdump 
    /usr/lib/rpm/brp-python-bytecompile 
    /usr/lib/rpm/redhat/brp-python-hardlink 
    /usr/lib/rpm/redhat/brp-java-repack-jars 

F10 (i386):

    /usr/lib/rpm/brp-compress
    /usr/lib/rpm/brp-strip
    /usr/lib/rpm/brp-strip-static-archive
    /usr/lib/rpm/brp-strip-comment-note

F10 (i386) -- in an i386 mock chroot:

    /usr/lib/rpm/redhat/brp-compress
    /usr/lib/rpm/redhat/brp-strip /usr/bin/strip
    /usr/lib/rpm/redhat/brp-strip-static-archive /usr/bin/strip
    /usr/lib/rpm/redhat/brp-strip-comment-note /usr/bin/strip /usr/bin/
objdump
    /usr/lib/rpm/brp-python-bytecompile
    /usr/lib/rpm/redhat/brp-python-hardlink
    /usr/lib/rpm/redhat/brp-java-repack-jars

F11 (i386):

    /usr/lib/rpm/brp-compress
    /usr/lib/rpm/brp-strip
    /usr/lib/rpm/brp-strip-static-archive
    /usr/lib/rpm/brp-strip-comment-note

F11 (i386) -- in an i386 mock chroot

    /usr/lib/rpm/redhat/brp-compress
    /usr/lib/rpm/redhat/brp-strip /usr/bin/strip
    /usr/lib/rpm/redhat/brp-strip-static-archive /usr/bin/strip
    /usr/lib/rpm/redhat/brp-strip-comment-note /usr/bin/strip /usr/bin/
objdump
    /usr/lib/rpm/brp-python-bytecompile
    /usr/lib/rpm/redhat/brp-python-hardlink
    /usr/lib/rpm/redhat/brp-java-repack-jars

Why does the %__os_install_post macro vary in these examples?  It's a 
little frustrating, since a spec file that works outside of a mock chroot 
may not work inside, and vice-versa.

C

[Jeremy Katz]

On Tuesday, April 28 2009, Carl D. Roth said:
>  I noticed this wierdness with rpm-4.6 (f10) and rpm-4.7 (f11)... 
> When I 
>  build packages including Python scripts, sometimes the .pyc and .pyo 
>  files are included in the package manifest, and sometimes, they are not. 
[snip] 
>  Why does the %__os_install_post macro vary in these examples? 
> It's a 
>  little frustrating, since a spec file that works outside of a mock chroot 
>  may not work inside, and vice-versa. 
Make sure you always have redhat-rpm-config installed

Jeremy

[Carl D. Roth]

On Tue, 28 Apr 2009 11:36:20 -0400, Jeremy Katz wrote:

>  On Tuesday, April 28 2009, Carl D. Roth said:
> > I noticed this wierdness with rpm-4.6 (f10) and rpm-4.7 (f11)...  When
> > I build packages including Python scripts, sometimes the .pyc and .pyo
> > files are included in the package manifest, and sometimes, they are
> > not.
>  [snip]
> > Why does the %__os_install_post macro vary in these examples?  It's a
> > little frustrating, since a spec file that works outside of a mock
> > chroot may not work inside, and vice-versa.
>  
>  Make sure you always have redhat-rpm-config installed
>  
>  Jeremy 
Since it appears to be necessary to build RPMs on a Fedora system, why 
isn't there a dependency chain between redhat-rpm-config and rpm-build?

C

[Panu Matilainen]

On Wed, 29 Apr 2009, Carl D. Roth wrote:

>  On Tue, 28 Apr 2009 11:36:20 -0400, Jeremy Katz wrote:
>
> > On Tuesday, April 28 2009, Carl D. Roth said:
> >> I noticed this wierdness with rpm-4.6 (f10) and rpm-4.7 (f11)...  When
> >> I build packages including Python scripts, sometimes the .pyc and .pyo
> >> files are included in the package manifest, and sometimes, they are
> >> not.
> > [snip]
> >> Why does the %__os_install_post macro vary in these examples?  It's a
> >> little frustrating, since a spec file that works outside of a mock
> >> chroot may not work inside, and vice-versa.
> >
> > Make sure you always have redhat-rpm-config installed
> >
> > Jeremy
>
>  Since it appears to be necessary to build RPMs on a Fedora system, why
>  isn't there a dependency chain between redhat-rpm-config and rpm-build? 
redhat-rpm-config defines RH/Fedora build policy/configuration. Rpm itself 
doesn't give a damn where that configuration comes from and is fully 
functional without it. It's just that rpm upstream build "policy"
doesn't 
match Fedora's.

 	- Panu -

[Kevin Kofler]

Panu Matilainen wrote:
>  redhat-rpm-config defines RH/Fedora build policy/configuration. Rpm
> itself
>  doesn't give a damn where that configuration comes from and is fully
>  functional without it. It's just that rpm upstream build "policy"
> doesn't
>  match Fedora's. 
But Fedora's rpm(-build) RPM should really require redhat-rpm-config.

        Kevin Kofler

[Panu Matilainen]

On Thu, 30 Apr 2009, Kevin Kofler wrote:

>  Panu Matilainen wrote:
> > redhat-rpm-config defines RH/Fedora build policy/configuration. Rpm itself
> > doesn't give a damn where that configuration comes from and is fully
> > functional without it. It's just that rpm upstream build "policy"
> doesn't
> > match Fedora's.
>
>  But Fedora's rpm(-build) RPM should really require redhat-rpm-config. 
Well, redhat-rpm-config is required to build "Fedora compliant" packages 
(whatever that means) but it is not required to build packages generally.

A semi-reasonable middle-ground might be adding a virtual provide such as 
"system-rpm-config" to redhat-rpm-config and have rpm-build depend on 
that. It's not correct either as rpmbuild is functional without extra 
configuration, but it at least leaves users the option of replacing the 
Fedora build policy with something else.

 	- Panu -

[James Antill]

On Thu, 2009-04-30 at 10:41 +0300, Panu Matilainen wrote:
>  On Thu, 30 Apr 2009, Kevin Kofler wrote:
>  
>  > Panu Matilainen wrote:
>  >> redhat-rpm-config defines RH/Fedora build policy/configuration. Rpm itself
>  >> doesn't give a damn where that configuration comes from and is fully
>  >> functional without it. It's just that rpm upstream build "policy"
> doesn't
>  >> match Fedora's.
>  >
>  > But Fedora's rpm(-build) RPM should really require redhat-rpm-config.
>  
>  Well, redhat-rpm-config is required to build "Fedora compliant" packages 
>  (whatever that means) but it is not required to build packages generally.
>  
>  A semi-reasonable middle-ground might be adding a virtual provide such as 
>  "system-rpm-config" to redhat-rpm-config and have rpm-build depend on 
>  that. It's not correct either as rpmbuild is functional without extra 
>  configuration, but it at least leaves users the option of replacing the 
>  Fedora build policy with something else. 
 Another option might be to have fedora-packager require
redhat-rpm-config, which is closer to the correct requires and probably
what people expect.

-- 
James Antill <james(a)fedoraproject.org&gt;
Fedora