On Sat, Dec 18, 2004 at 08:40:41AM +0100, Eric Botcazou wrote: > Because invoking again resolve_unique_section on the decl will be a no-op. Yes... > Now the second invocation takes into account the 'reloc' parameter ... Which will always be zero, because the symbol is uninitialized, otherwise we wouldn't have gone through the first resolution. r~