Hi all,
I have started to work with c++ and autotools.
I mainly maintain an application and a shared library as a hobby.
Nothing professional yet, because I want to understand the best practice strongly recommended by advanced professional programmers.

I check out my sources into $HOME/repository and svn creates:
$HOME/repository/app/application/src
$HOME/repository/lib/libutil/src
where I get c++ sources/headers + configure.ac and Makefile.am

Now I need to understand what is considered as a good practice for debugging application, either A), B) or something else.

A) install application and its dependency with:
./configure --prefix=$HOME/local
./make
./make install
export LD_LIBRARY_PATH=$HOME/local/lib
and debug application from this setup [$HOME/local/bin]

or

B)
using AM_LDFLAGS=-rpath $HOME/repository/libutil/src/.libs in Makefile.am to load a dependency from a hardcoded path.
In this way I do only:
./configure
./make

and then I debug the application in the place it was built. [$HOME/repository/app/src/]
This is the current way I am doing it.
Please understand I:
-never distributed my application
-never installed it system wide.

Can a programmer with more experience tell me more about this ?
Thank you.

As far as the generic question of whether you should test your binary in the source directory or after install: it doesn't really matter, as long as you know the binary will work in both places, and it is probably more convenient to test it without having to run "make install" first every time.

Down the road, if your application starts using external resources (image files, music, etc.) you may end up having to run make install before testing, in order to get new or modified resources into the place where the program expects to find them. This is something you'll want to keep in the back of your mind as you program. There are various ways of handling that issue, but one way is to have the configure script hard code the data installation directory into the config.h file when it is run, in which case the program will have to be installed in order to run correctly.

As far as the AM_LDFLAGS=-rpath... part, I haven't every used that myself, but I found this link that mentions both positive and negative reactions:

http://www.sourceware.org/autobook/a...tobook_88.html

For what it's worth:

In general, I like to keep any debugging/test versions of software as close as possible to the "real" versions, so I'd go with the first option. Not that it's a huge difference at this stage, though, but it might be later as CoderMan pointed out.

And if you want to minimize typing, you can always create a wrapper script that does the build, installation & sets up LD_LIBRARY_PATH for you - especially useful if you've got lots of dependencies, or if you put lots of configure options in and you want to make sure your library at least builds & runs some test programs correctly with various combinations of them.

So, like CoderMan said - not much difference now, but maybe there will be in the future

P.S. Don't forget that build systems need debugging, too - it's nice to know that everything gets installed in its right place as early as possible, so you don't have to cope with it when you've got a large project and lots of errors related to the install!