From: Karsten Pedersen <kpedersen@gmx.com>
Subject: Absolute rpaths in some packages
To: ports@openbsd.org
Date: Thu, 18 Jul 2024 18:59:49 +0200

Hello there,

I have been developing some software[1] (effectively a hack) that
leverages the OpenBSD packages and creates standalone self contained
prefixes / "bundles". Particularly useful for behemoths such as
Firefox, Chromium, etc. In terms of updates, these guys drag in so
much cruft, I kind of just prefer to blow them away and start from
scratch anyway with a new i.e: /opt/chromium122

Through many of my cludges (mainly environment variables avoiding
hard coded /usr/local paths but also some binary patching of
rpaths[2]), I notice that a few libraries (and executables) built
by ports have absolute rpaths. But this is rare suggesting it isn't
really intended to be the norm. Is this something that should be
reported to the respective port maintainer? Examples are (objdump
-x [...] | grep NEEDED):

libtiff.so.42.0 (from the tiff package).
  /usr/local/lib/libzstd.so.6.3

xpdf (from the xpdf package (not xpdf3))
  /usr/local/lib/qt5/./libQt5Network.so.4.0
  /usr/local/lib/qt5/./libQt5PrintSupport.so.3.0
  /usr/local/lib/qt5/./libQt5Widgets.so.4.1
  /usr/local/lib/qt5/./libQt5Gui.so.4.2
  /usr/local/lib/qt5/./libQt5Core.so.6.0

Many thanks,

Karsten

[1] https://codeberg.org/kpedersen/pkg_bundle
[2] https://codeberg.org/kpedersen/pkg_bundle/src/branch/master/share/pkg_bundle/clean-rpath.c