I've done some debugging this afternoon and found that the problem is
due to my spec having a %clean section.
%install contains:
rm -rf %{buildroot}
make install DESTDIR=%{buildroot}
%clean contains:
rm -rf %{buildroot}
When "Executing(%install)" happens the temporary script in /var/tmp contains:
definition of env variables, e.g. RPM_SOURCE_DIR
commands from %install
exit
command from %clean
/usr/lib/rpm/check-buildroot, /usr/lib/rpm/redhat/brp-compress, etc.
The 'exit' means that the script exits early, and the man page
compression (amongst other things) doesn't happen.
If I remove the %clean section, the temporary script contains:
definition of env variables, e.g. RPM_SOURCE_DIR
commands from %install
/usr/lib/rpm/check-buildroot, /usr/lib/rpm/redhat/brp-compress, etc.
i.e. no "exit" and no "rm -rf" from %clean.
I've read this:
https://fedoraproject.org/wiki/EPEL:Packaging#Previously_required_boilerplate
so I understand %clean is redundant, and I will remove it - it solves
my problem. I also see the warning about not putting %clean
immediately after %description because its contents might get appended
to %description. Is that what's happening here? Is %clean being
incorrectly appended to %install?
I also had a look at the macros:
https://pkgs.fedoraproject.org/cgit/rpms/epel-rpm-macros.git/tree/macros.zzz-epel?h=el5
and noticed on line 200:
rpm.define("clean exit")
...which might explain why "%clean" gets turned into "exit" in the
temporary script.
So I think I've solved my problem, but I wonder if there's a bug in
the macros, or if the documentation needs to be improved slightly? For
me it seems that having a %clean section caused problems, even though
it wasn't "the first section after %description" as warned about on
the wiki.
Regards,
Richard
On 3 September 2016 at 14:58, Richard Fearn <richardfearn@xxxxxxxxx> wrote:
> Hi,
>
> Has something changed with the way that EPEL5 packages are built that
> means man pages are no longer automatically compressed?
>
> I've just tried to build a new ncdu package (version 1.12) for EPEL5,
> but it failed:
>
> http://koji.fedoraproject.org/koji/taskinfo?taskID=15482900
>
> The man page is installed uncompressed:
>
>> /usr/bin/install -c -m 644 ncdu.1 '/var/tmp/ncdu-1.12-1.el5-root-mockbuild/usr/share/man/man1'
>
> The build failed because ncdu.1.gz (which is meant to go into the
> RPMs) couldn't be found, and ncdu.1 (which shouldn't exist when the
> RPMs are created) was unpackaged.
>
> This wasn't an issue with the last build (1.11) I did in April 2015:
>
> http://koji.fedoraproject.org/koji/buildinfo?buildID=626540
>
> Again the man page was installed uncompressed, but had been
> (automatically) compressed by the time the RPM was created.
>
> Comparing the build logs, these lines from the 1.11 build aren't
> present for the 1.12 build:
>
>> + /usr/lib/rpm/find-debuginfo.sh /builddir/build/BUILD/ncdu-1.11
>> extracting debug info from /var/tmp/ncdu-1.11-1.el5-root-mockbuild/usr/bin/ncdu
>> 278 blocks
>> + /usr/lib/rpm/check-buildroot
>> + /usr/lib/rpm/redhat/brp-compress
>> + /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-java-repack-jars
>
> Thanks for any help!
>
> Rich
>
> --
> Richard Fearn
> richardfearn@xxxxxxxxx
--
Richard Fearn
richardfearn@xxxxxxxxx
_______________________________________________
epel-devel mailing list
epel-devel@xxxxxxxxxxxxxxxxxxxxxxx
https://lists.fedoraproject.org/admin/lists/epel-devel@xxxxxxxxxxxxxxxxxxxxxxx
