On Sun, Jul 30, 2017 at 12:52 PM, Sven Joachim <[email protected]> wrote: > Really? There seems to be quite a bit of code that's conditional to > #ifdev HAVE_VALGRIND, e.g. in intel/intel_bufmgr_gem.c, and I don't > think building with the nocheck profile should alter the produced binary > packages.
Hmm, I guess if it's indeed the case that it's generating valgrind support stub assembly, or something along those lines, then it might get more complex. It might be necessary to produce libdrm2-stage1 etc. packages with shlibs/symbols set up to generate dependencies on "libdrm2 | libdrm2-stage1". Either that, or break the cycle somewhere else, for example maybe by having valgrind able to produce a valgrind-stage1 package without the correct gdb path encoded in. -- Daniel
