stage3/std.o: $(STDLIB_CRATE) $(STDLIB_INPUTS) \ stage2/rustc$(X) stage2/$(CFG_STDLIB) stage2/intrinsics.bc \ $(LREQ) $(MKFILES) @$(call E, compile: $@) $(STAGE2) -c --shared -o $@ $< stage3/$(CFG_STDLIB): stage3/std.o stage3/glue.o @$(call E, link: $@) $(Q)gcc $(CFG_GCC_CFLAGS) stage3/glue.o $(CFG_GCC_LINK_FLAGS) -o $@ $< \ -Lstage3 -Lrt -lrustrt stage3/rustc.o: $(COMPILER_CRATE) $(COMPILER_INPUTS) $(SREQ2) @$(call E, compile: $@) $(STAGE2) -c -o $@ $< stage3/glue.o: stage2/rustc$(X) stage2/$(CFG_STDLIB) stage2/intrinsics.bc \ rustllvm/$(CFG_RUSTLLVM) rt/$(CFG_RUNTIME) @$(call E, generate: $@) $(STAGE2) -c -o $@ --glue stage3/intrinsics.bc: $(INTRINSICS_BC) @$(call E, cp: $@) $(Q)cp $< $@ # Due to make not wanting to run the same implicit rules twice on the same # rule tree (implicit-rule recursion prevention, see "Chains of Implicit # Rules" in GNU Make manual) we have to re-state the %.o and %.s patterns here # for different directories, to handle cases where (say) a test relies on a # compiler that relies on a .o file. stage3/%.o: stage3/%.s @$(call E, assemble [gcc]: $@) $(Q)gcc $(CFG_GCC_CFLAGS) -o $@ -c $< stage3/%$(X): stage3/%.o $(SREQ2) @$(call E, link [gcc]: $@) $(Q)gcc $(CFG_GCC_CFLAGS) stage3/glue.o -o $@ $< \ -Lstage3 -Lrustllvm -Lrt -lrustrt -lrustllvm -lstd -lm @# dsymutil sometimes fails or prints a warning, but the @# program still runs. Since it simplifies debugging other @# programs, I\'ll live with the noise. -$(Q)$(CFG_DSYMUTIL) $@