Re: Building rpms in mock

> I don't have any immediate plans to add more configure options, 
> but still I don't like how this approach ends up duplicating the 
> list of options in the spec file...

I agree. I already called it "messy" and "ugly". And you certainly
have a point in that

> rpm packages really are not meant to offer custom configuration 
> options, but to build a known package that's configured in a 
> known way, for a distribution. 

But even so, there are some other problems.

> A courier-authlib-config-%{mailuser}-%{mailgroup} empty subpackage. 
> So, it'll be courier-authlib-config-daemon-daemon for me, and 
> courier-authlib-config-courier-courier for you. [...]  Both 
> courier-authlib, and all other packages will get a dependency on 
> this subpackage.
This narrows down the possible user:group names to courier and
daemon and it's good, though anyone currently running as something
else, e.g. mail:daemon, will be left out in the cold.

The main problem I'm facing is that neither authlib nor courier
can be built in mock because they expect the user:group to exist.
So I wonder, why is that check necessary? If a --define has
already told configure what user:group to use, why can't the
rpm just go ahead and build with that user:group even if they
do not exist on the build host? And conversely, if configure
has not been told what user:group to use, why can't the rpm
just go ahead and build with the default user:group? After all,
even if someone builds an rpm with mismatching user:group for
an upgrade of his existing installation, the authlib-config
metapackage will now prevent him from installing it.

As for the installation host, since the rpm knows what user:group
to use, it's a simple matter to check for them in a %pre and create
them if need be. BTW, contrary to what I did in my previous %pre,
useradd -d should be %{prefix} instead of /usr/lib/courier so it
gets correctly expanded if someone changes the preset prefix.

Another problem comes with the old-style system detections by
fedora-release, rocky-release etc. A simple %{?fedora} would
cover the former and a %{?rhel} would cover all el variants and
derivatives, including alma which I am using and the rpm now
identifies as non-fedora and non-rhel. (I don't know whether
centos-stream identifies itself as rhel, fedora, or something
else.)

Finally, on line 170 of the 1.3.5-1 courier.spec there is a
dangling "%if %is_rocky" that I don't understand. But at least
this can be done regarding fedora:

--- rpmbuild/SPECS/courier.spec 2023-12-26 10:53:07.000000000 +0000
+++ rpmbuild/SPECS/courier.spec 2023-12-26 10:58:36.931000000 +0000
@@ -146,11 +146,6 @@
  BuildRequires:   perl(ExtUtils::Embed)
  BuildRequires:   /etc/mime.types

-%define need_perl_generators %(if rpm -q fedora-release >/dev/null 2>/dev/null; then echo "1"; exit 0; fi; echo "0"; exit 1)
-
-%if %need_perl_generators
-BuildRequires: perl-generators
-%endif

  %define use_openssl %(rpm -q openssl-devel >/dev/null 2>&1 && echo 1 && exit 0; echo 0)

@@ -178,10 +173,9 @@
  BuildRequires:      %{__make}
  Obsoletes:     courier-smtpauth < 1.3.5

-%define is_fedora %(if rpm -q fedora-release >/dev/null 2>/dev/null; then echo "1"; exit 0; fi; echo "0"; exit 1)
-
-%if %is_fedora
+%if %{?fedora}
  BuildRequires: /usr/lib/locale/locale-archive
+BuildRequires: perl-generators
  %endif

  %if ! %alternatives

perl-generators are also needed by all rhel. I assume that
the same applies to /usr/lib/locale/locale-archive; up to
el7 this came from glibc-common and from el8 on it is part
of glibc-all-langpacks. If so, %if ( %{?fedora} || %{?rhel} )
would do the trick for both at the same time.

Cheers,

Z


-- 
Слава Україні!
Путин хуйло!


_______________________________________________
courier-users mailing list
[email protected]
Unsubscribe: https://lists.sourceforge.net/lists/listinfo/courier-users