Bugzilla Bug 33106

libpq++-4.0-r1.ebuild does odd tricks to fixup installation paths for libpq++.
The attached .ebuild replaces those tricks by a patch, which allows passing
variables to the make program.

The .ebuild also changes the installation locations for libpq++: The current
.ebuild installs in in /usr/include and /usr/lib, whilest all other postgres
files are in /usr/include/postgres and /usr/lib/postgres. From reading libpq++'s
Makefile it appears like the upstream maintainer also suggests
$prefix/include/postgres and $prefix/lib/postgres. This this behavoir also
reduces /usr/include pollution a little bit, I've adopted the second schema.
This step should not cause any problems, since linking with libpq++ also
requires you to like with libpq, which requires /usr/include/postgres to be
added to the include-path and /usr/lib/postgres to the library path anyway.

Well, and finally I've added some statements to install libpq++'s documentation.
Something that's very important for developers using the library. ;-) 

Is there a way to avoid the stuff in examples/ getting compressed?
Should the stuff in examples/ and html/ only be installed, when "docs" is in USE?

Reproducible: Always
Steps to Reproduce:

Created an attachment (id=20486) [edit]
The new .ebuild

Created an attachment (id=20487) [edit]
The directory-variable patch.

Forgot to mention: Also updated HOMEPAGE to point on
http://gborg.postgresql.org/project/libpqpp/projdisplay.php.
Well, and accordingly to that page the package's license is BSD, not LGPL-2.1.

Created an attachment (id=20490) [edit]
A new new .ebuild

D'uh: Just looked at libpq. Installing any crap in /usr/include and /usr/lib
appears to be common sense among the Gentoo core developers. Well, don't
like
it, but I've modified my .ebuild to follow that convention.

Created an attachment (id=20491) [edit]
A new directory-variable-patch

This new patch additionally fixes the way the Makefile creates the libpq++.so
symlink: The Makefile tries to create an absolute symlink, which must produce
nonsense within the sandbox. The patch turns the symlink into a relative
symlink.

The license is still wrong in all dev-cpp/libpqpp ebuilds. Otherwise, I have no
clue if this bug even remotely applies now. So - pgsql herd, please fix the
license at least. 

Finally fixed. LOL, it took more than 3 years to fix this!!!