I recently posted a "my bad", but it lead me to wonder why "make test" was failing. I had commit 64f8434d0 from 02-August-2013 installed, and had built commit 9be83a4d from 16-August-2013 and was executing "make test" from that build directory. Tests were failing because a symbol was missing from a library that, according to the debug output, should not have been being used.
So, today I spent some time playing around with GrTest.cmake and the settings for GR_TEST_TARGET_DEPS. For this test, I deactivated -all- installed versions of GNU Radio. I then built (but did not install) 9be83a4d from 16-August-2013. "make test" failed again for (almost?) all tests because libraries were not found. After poking around, I see that in the top-level CMakeLists.txt file, GR_TEST_TARGET_DEPS is being set to "volk gnuradio-runtime". Then, in many of the components, I find 'set(GR_TEST_TARGET_DEPS "")'! Which says to me that this component has no internal dependencies; which is not going to be the case except for maybe gruel or pmt. And, the cmake test shell scripts reflect this too (no internal dependencies). Some of the components have actual dependencies set, e.g., gr-filter: 'list(APPEND GR_TEST_TARGET_DEPS test-gr-filter gnuradio-filter gnuradio-fft)'. But, again, some of the required libraries are still not being found. I believe that most of the GR_TEST_TARGET_DEPS settings are not correct, and, further, that in the top-level setting gnuradio-pmt should be added. I'm happy to hack through this and figure out what needs to set and to what. But, I'd prefer a sanity check from others (Linux or OSX) that shows that my beliefs are founded. So, can some other folks try -building- (not installing) GNU Radio from the latest (master or next) with no other GNU Radio installed, and see if "make test" works? Thanks! - MLD _______________________________________________ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org https://lists.gnu.org/mailman/listinfo/discuss-gnuradio