Hi all,
I have a problem that's beyond my limited RPM knowledge and need some
advice.
I have a source tree for software that runs on a small cluster. There are
several different node types, and each type requires different software. The
idea is to create one binary RPM per node type, then use the cluster tools
to distribute and install the RPMs to their correct
nodes.
My 'make install' target creates a directory structure as follows:
$INSTALL_ROOT/node1/etc/...
$INSTALL_ROOT/node1/bin/...
$INSTALL_ROOT/node1/opt/...
$INSTALL_ROOT/node2/etc/...
$INSTALL_ROOT/node1/bin/...
$INSTALL_ROOT/node2/opt/...
$INSTALL_ROOT/node3/etc/...
$INSTALL_ROOT/node1/bin/...
$INSTALL_ROOT/node3/opt/...
My spec file looks something like this (some parts omitted):
Summary: Cluster software
Name: cluster
Version: 1
Release: 1
---snip---
%package node1
Summary: Cluster software for node 1
%description
Software for node 1
%package node2
Summary: Cluster software for node 2
%description
Software for node 2
%package node3
Summary: Cluster software for node 3
%description
Software for node 3
---snip---
%install
make install INSTALL_ROOT=$RPM_BUILD_ROOT
%files node1
%defattr(-,root,root)
/node1/*
%files node2
%defattr(-,root,root)
/node2/*
%files node3
%defattr(-,root,root)
/node3/*
The problem is the files in the binary packages obviously have /node1,
/node2, etc prefixes. Is there a way to strip these off so the packages
install as intended?
I've thought about not using the prefixes at all in 'make install'. But
there are config files with identical names that must be installed on each
node.
I might be going about this the wrong way... if so, I'd really appreciate
suggestions. I *can* change the way 'make install' works.
Thanks in advance,
Jim
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
