What do you keep in /usr/local/src?

I keep the tarballs so they can't be damaged... then if I want to install something I tar -zxvf /usr/local/src into a temporary folder...

Do you create a seperate build folder? What do you do with the remains? How do you preserve the Makefile (and its dependencies?)

I am looking at keeping all the Makefiles in /usr/local/uninstall/[programfolder]

Is it better to just keep the folders how you built them in /usr/local/src? You must build somewhere else first then (because you don't want to do ./configure; make as root(?).. then move it to /usr/local/src?

How do you install things from source, and where do you keep stuff?

This is assuming you don't use any manager such as CheckInstall...

I just have a download directory in my $HOME where all the archives go, somewhat organized according to what sort of program it is. I don't care about saving makefiles...if anything needs special attention beyond configure/make/make install then I'll either make a text file to remind me or if the README and INSTALL files are good enough I leave it at that.

As far as dependencies, I usually keep those with the main application's tarball when it's something i wouldn't think to install otherwise.

I was leaning towards a "more than just yourself" using the computer... so nothing really goes in the ~/ dir

I create a /usr/local/src/.tarballs directory. It keeps the tarballs there, and removes the directory from sight when doing listings. Similarly, patches either go in the .tarball directory or a .patches directory; whichever suits your method of organization.

I "cd" to /usr local/src and issue command like: tar xvjf .tarballs/tarball_name.tar.bz2

I build within the directory itself unless specified to do otherwise in the docs.

I store a file with the commands used to compile and install the source. This is usually just a list of the patch commands used and the options to configure; make and make install rarely change.

I run two installs. The first is a compile and install with a non-privileged user. and to a non-standard location. For instance, installing to this user's home directory (something like /home/testinst/local_install). Then I run a find command on the directory to get a list of all the files created/copied, and then make uninstall or just simply delete the whole local_install subtree. The second install is the "real" install, with the location being set to /usr/local or wherever.

After installing, I remove the source tree and save the two files I created: the one with commands to patch/configure the source and the one listing the files installed in some other location (again, wherever your organizational preferences tell you). This way, I can always re-extract the source and get the source tree back to exactly the state it was when I compiled and installed without the need for keeping the source tree around forever. It also allows for manual uninstall if the developers left out that target in the makefile.

Pitfalls of this method: If the install script is hard-wired to copy something to a system location (such as /usr/lib), then its possible the file listing isn't complete. That is, if the make install on the /home/testint/local_install fails because it doesn't have root privileges to make the copy, then commands following that copy may not get executed. Similarly, if make install modified any configuration files, this method will not show it. There might be more, but I haven't noticed them yet.