Spurious R_386_NONE relocations

[Jeff Baker]

I have a test project that, when compiled with gcc 2.95.3 and the latest

binutils, ends up with a load of R_386_NONE relocations.  I've tried
tweaking our loader to ignore them, but the program crashes if you do
this so I'm under the impression that they are the result of bug.  They
appear exclusively in the exception handling sections of a c++ shared
object.

I know that's pretty vague but does this sound familiar to anyone?  The
problem doesn't exist with gcc 3.3.1+ but it looks like it's because the

compiler is outputting different code, not because the 'bug' has been
fixed in the linker.