Bug 2386540

No matching package to install: 'gnuplot'
No matching package to install: 'rsh'

gnuplot has been requested.  I think it's probably time to figure out how to drop the rsh dep.

I can have look at gnuplot in about one week. I think that the rsh dep is silly, as rsh can be usually replaced by ssh 1:1. It has been a long time since I looked how Amanda works, though.

By the way, the gnuplot dependency is also not critical. Here is an old patch I used to reduce it:

diff --git a/amanda.spec b/amanda.spec
index 114b3c5..7b51476 100644
--- a/amanda.spec
+++ b/amanda.spec
@@ -1,3 +1,9 @@
+%if 0%{?rhel} >= 9
+%bcond_with gnuplot
+%else
+%bcond_without gnuplot
+%endif
+
 %define _libexecdir %{_libdir}
 %define defconfig DailySet1
 %define indexserver amandahost
@@ -10,7 +16,7 @@
 Summary:       A network-capable tape backup solution
 Name:          amanda
 Version:       3.5.1
-Release:       27%{?dist}
+Release:       28%{?dist}
 Source:        http://downloads.sourceforge.net/amanda/amanda-%{version}.tar.gz
 Source1:       amanda.crontab
 Source4:       disklist
@@ -42,7 +48,10 @@ License:       BSD and GPLv3+ and GPLv2+ and GPLv2
 URL:           http://www.amanda.org
 BuildRequires: automake autoconf libtool
 BuildRequires: xfsdump
-BuildRequires: gnuplot cups samba-client tar grep
+%if %{with gnuplot}
+BuildRequires: gnuplot
+%endif
+BuildRequires: cups samba-client tar grep
 BuildRequires: gcc-c++ readline-devel libtirpc-devel
 BuildRequires: krb5-devel rsh openssh-clients ncompress mtx mt-st
 BuildRequires: perl-devel perl-generators perl(ExtUtils::Embed) perl(Test::Simple)
@@ -138,6 +147,9 @@ export MAILER=/bin/mail CFLAGS="$RPM_OPT_FLAGS -fPIE" LDFLAGS="$RPM_LD_FLAGS -pi
            --with-group=%amanda_group \
            --with-tmpdir=/var/log/amanda \
            --with-gnutar=/bin/tar \
+%if %{without gnuplot}
+           --without-gnuplot \
+%endif
            --with-ssh-security \
            --with-rsh-security \
            --with-bsdtcp-security \
@@ -227,11 +239,13 @@ make check
 %{_libexecdir}/amanda/amandad
 %{_libexecdir}/amanda/amanda-sh-lib.sh
 %attr(4750,root,%amanda_group) %{_libexecdir}/amanda/ambind
-%{_libexecdir}/amanda/amcat.awk
 %{_libexecdir}/amanda/amndmjob
+%if %{with gnuplot}
+%{_libexecdir}/amanda/amcat.awk
 %{_libexecdir}/amanda/amplot.awk
 %{_libexecdir}/amanda/amplot.g
 %{_libexecdir}/amanda/amplot.gp
+%endif
 %{_libexecdir}/amanda/ndmjob
 
 %{_sbindir}/amandad
@@ -243,7 +257,9 @@ make check
 %{_sbindir}/amcryptsimple
 %{_sbindir}/amgetconf
 %{_sbindir}/amgpgcrypt
+%if %{with gnuplot}
 %{_sbindir}/amplot
+%endif
 %{_sbindir}/amssl
 
 %{perl_vendorarch}/Amanda/
@@ -257,7 +273,9 @@ make check
 %{_mandir}/man7/amanda-scripts.7*
 %{_mandir}/man8/amanda.8*
 %{_mandir}/man8/amarchiver.8*
+%if %{with gnuplot}
 %{_mandir}/man8/amplot.8*
+%endif
 %{_mandir}/man8/script-email.8*
 %{_mandir}/man8/amaespipe.8*
 %{_mandir}/man8/amcrypt-ossl-asym.8*
@@ -441,6 +459,9 @@ make check
 
 
 %changelog
+* Tue Feb  9 2021 Pavel Cahyna <pcahyna> - 3.5.1-28
+- Make the gnuplot dependency together with building of amplot optional.
+
 * Tue Jan 26 2021 Fedora Release Engineering <releng> - 3.5.1-27
 - Rebuilt for https://fedoraproject.org/wiki/Fedora_34_Mass_Rebuild

It turns out I have a need for this myself.

I agree about rsh; it should just be a matter of removing --with-rsh-security from the configure line and I'm thinking that I'll just do that in rawhide.  I can't imagine that at this point anyone could be relying on rsh, but maybe --with-rsh-security has some other side effect that I don't know about.  Gnuplot is already conditionalized, and indeed Pavel's patch was already applied (by him...) in 2021 so that shouldn't be an issue.

So I do now have a functional set of packages for Almalinux 10.  All that's needed is to remove the dependency on rsh and the --with-rsh-security flag, as well as fixing one unfortunate issue I introduced with a recent comment that beaks on machines that don't have merged bin and sbin.

I can only test the client side in my environment.

The problem that prevents me from just pushing this is that the client really seems to want an /etc/dumpdates file to be present and writeable.  Normally this would come from the dump package, but of course that no longer exists; personally I use xfsdump and tar.  (And sadly tar has its own issues).

I haven't spent any time digging into what it would take to get the client code to stop complaining about this file.  If anyone knows, feel free to save me some time.  Otherwise I'll try to dig into it in the next few weeks.