I'm writing a build process using a Perl script thats primary function is to create a chroot() jail before running "make install" so that a package can install itself into what it *thinks* is the root filesystem but is actually another location altogether (later I can tar up the results and release the tarball onto other machines for release).

e.g. I create a directory /tmp/install, and have my source in /tmp/build. My script does the following:

• chdir() into /tmp/build. Run configure and make
• runs mount of /bin, /usr/bin, /lib, and /tmp/build into /tmp/install/bin, /tmp/install/usr/bin, /tmp/install/lib, and /tmp/install/tmp/build respectively using the --bind mount option
• chroot() into /tmp/install
• run make install inside the chroot jail
• umount the bin, usr/bin, lib, and tmp/build directories

I can tar up the contents of /tmp/install and have a clean build that can be applied to the root file system of similar machines.

My problem is that I want to build mod_perl and this a)needs Perl and b)wants to install things into the Perl library directories. A simple mount of the Perl directory is not sufficient because the Perl modules installed will go straight into my system (and I don't want that).

So I want to copy all of /usr/local/perl into /tmp/install/usr/local/perl, run my chroot(), do the make install, remove all the copied files from /tmp/install/usr/local/perl, and then be left with the Perl modules that were created in /tmp/install/usr/local/perl/lib/....

I thought about maybe using File::Find and File::Copy along with File::Path to automatically create directories, and storing each copied file into an array for subsequent deletion. Is this the best approach? How have other people applied a directory copy then removal (leaving any new artifacts in place).

update: corrected spelling.