Date: Wed, 13 Jun 2012 00:05:20 +0200 From: Matthieu Volat <mazhe@alkumuna.eu> To: ports@freebsd.org Subject: Re: Imagemagick: FAIL, now in Magick++/demo/analyze.sh Message-ID: <20120613000520.12015a65.mazhe@alkumuna.eu>
next in thread | raw e-mail | index | archive | help
On Sat, Jun 9, 2012 at 8:02 PM, Doug Barton <dougb at freebsd.org> wrote: > The new version fails in a new location, even with the default options. > It seems to me that this failure cames from the fact that the tests try to dynamicaly load modules from the path where the previous version is installed instead of build dirs. Basicaly, the test do the following (I've cut through various shell wrappers to be able to use gdb): cd /usr/ports/graphics/ImageMagick/work/ImageMagick-6.7.7-7 env LD_LIBRARY_PATH=magick/.libs:wand/.libs:Magick++/lib/.libs MAGICK_CODER_MODULE_PATH=coders/.libs MAGICK_FILTER_MODULE_PATH=filters/.libs ./Magick++/demo/.libs/analyze ./Magick++/demo/model.miff The problem is, you can check by running the test through gdb, that amongst the dlopen-ed libraries, /usr/local/lib/ImageMagick-6.7.7/modules-Q16/filters/analyze.so is used as if the MAGICK_FILTER_MODULE_PATH environment variable was not working. A quick and dirty test is to save /usr/local/lib/ImageMagick-6.7.7/modules-Q16/filters/analyze.so and /usr/local/lib/ImageMagick-6.7.7/modules-Q16/filters/analyze.la, and to replace them with the versions just built. In my case, the test passed. I am not too sure about how to fix that however... If I have some time, I will check that the MAGICK_FILTER_MODULE_PATH value is really checked before the normal installation path. About my setup : I'm running freebsd 9.0, amd64 and building with clang, openmp support is disabled. -- Matthieu Volat <mazhe@alkumuna.eu>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120613000520.12015a65.mazhe>