1 files changed, 50 insertions, 0 deletions
diff --git a/build-0.3/m4/m4.make b/build-0.3/m4/m4.make
new file mode 100644
index 0000000..ef5726b
--- /dev/null
+++ b/build-0.3/m4/m4.make
@@ -0,0 +1,50 @@
+# file      : build/m4/m4.make
+# copyright : Copyright (c) 2005-2012 Code Synthesis Tools CC
+# license   : GNU GPL v2; see accompanying LICENSE file
+
+$(out_base)/%: m4 := m4
+$(out_base)/%: m4_options +=
+
+ifeq ($(out_base),$(src_base))
+$(out_base)/%: $(src_base)/%.m4
+else
+$(out_base)/%: $(src_base)/%.m4 | $$(dir $$@).
+endif
+	$(call message,m4 $<,$(m4) $(m4_options) $< >$@)
+
+ifneq ($(out_base),$(src_base))
+
+$(out_base)/%: $(out_base)/%.m4 | $$(dir $$@).
+	$(call message,m4 $<,$(m4) $(m4_options) $< >$@)
+
+endif
+
+
+# @@
+# This is where things start breaking. Following standard logic I should
+# make a $(out_base)/%.clean rule, i.e., "will clean anything" rule. If
+# this rule happened to be before some other, more specialized rule, and
+# that rule happened to rm some additional stuff (like %.o tries to rm
+# .d file, which is also not quite correct...). In other word there
+# doesn't seem to be a way to properly match "build" and "clean" rules.
+# One idea is to make the "clean" rule depend on what "build" rule
+# depends (%.m4 in our case) hoping that this way the rule won't match.
+#
+# There are two problems with this approach:
+#
+# 1. It is if not iff. However, since the rules come in pairs and make
+#    pick the first implicit rule that matches, it is certain that if
+#    make picked this "clean" rule it also picked corresponding "build"
+#    rule.
+#
+# 2. The prerequisite (%.m4) can be an intermidiate file which itself
+#    may not exist. We don't want make to build it just to clean it
+#    or, even worse, to leave it laying around. I guess the only way
+#    to work around this is to provide special do-nothing rules during
+#    cleanup.
+#
+#
+.PHONY: $(out_base)/%.m4.clean
+
+$(out_base)/%.m4.clean:
+	$(call message,rm $(@:.m4.clean=),rm -f $(@:.m4.clean=))