Rémi Cardona wrote: > Andrew Haley a écrit : >> True. Either way, a test case will tell us. Speculation is pointless; >> let's see the code. > > While trying to slim down my current code to a comprehensive test-case, > I found out that my original bug can also be "fixed" by moving the > "utils_atom_get_property" function inside the same source file as the > one calling it. > > I made a dozen git branches in my code to test various ways of building > my code : static lib + program_LDFLAGS, static lib + program_LDADD, > libtool helper lib + program_LDADD, one big executable. All of them fail > identically. > > Only moving the actual function code to the calling source file solves > the bug. > > So, would that make it a linker bug? Maybe. > Are there ways to make sure that gcc is not part of the problem? Yes. Make me a self-contained reproducible test case and I'll tell you where the problem is. Andrew.
