From 9323067e91d8b26db1755bf9b72ccee9cf293464 Mon Sep 17 00:00:00 2001
From: Michael Martin <mcmartin@gmail.com>
Date: Sun, 3 Jun 2012 23:50:29 -0700
Subject: [PATCH] Improve .incbin to let its arguments be arbitrary expressions

This introduces a new IR node for mutable-during-assembly ranges.
In the common case, where offset and range are hardcoded or missing,
it continues with the older, more efficient behavior.
---
 src/Ophis/CorePragmas.py | 33 +++++++++++++++++++++++++--------
 src/Ophis/Passes.py      | 17 +++++++++++++++++
 2 files changed, 42 insertions(+), 8 deletions(-)

diff --git a/src/Ophis/CorePragmas.py b/src/Ophis/CorePragmas.py
index a6373eb..f6b25a1 100644
--- a/src/Ophis/CorePragmas.py
+++ b/src/Ophis/CorePragmas.py
@@ -53,26 +53,43 @@ def pragmaRequire(ppt, line, result):
 def pragmaIncbin(ppt, line, result):
     "Includes a binary file"
     filename = line.expect("STRING").value
-    offset = 0
-    size = -1
+    offset = IR.ConstantExpr(0)
+    size = None
     if str(line.lookahead(0)) == ",":
         line.pop()
-        offset = line.expect("NUM").value
+        offset = FE.parse_expr(line)
         if str(line.lookahead(0)) == ",":
             line.pop()
-            size = line.expect("NUM").value
+            size = FE.parse_expr(line)
     line.expect("EOL")
     if type(filename) == str:
         try:
             f = file(os.path.join(FE.context_directory, filename), "rb")
-            f.seek(offset)
-            bytes = f.read(size)
+            if offset.hardcoded and (size is None or size.hardcoded):
+                # We know how big it will be, we can just use the values.
+                if size is None:
+                    size = IR.ConstantExpr(-1)
+                f.seek(offset.value())
+                bytes = f.read(size.value())
+                bytes = [IR.ConstantExpr(ord(x)) for x in bytes]
+                result.append(IR.Node(ppt, "Byte", *bytes))
+            else:
+                # offset or length could change based on label placement.
+                # This seems like an unbelievably bad idea, but since we
+                # don't have constant prop it will happen for any symbolic
+                # alias. Don't use symbolic aliases when extracting tiny
+                # pieces out of humongous files, I guess.
+                bytes = f.read()
+                bytes = [IR.ConstantExpr(ord(x)) for x in bytes]
+                if size is None:
+                    size = IR.SequenceExpr([IR.ConstantExpr(len(bytes)),
+                                            "-",
+                                            offset])
+                result.append(IR.Node(ppt, "ByteRange", offset, size, *bytes))
             f.close()
         except IOError:
             Err.log("Could not read " + filename)
             return
-        bytes = [IR.ConstantExpr(ord(x)) for x in bytes]
-        result.append(IR.Node(ppt, "Byte", *bytes))
 
 
 def pragmaCharmap(ppt, line, result):
diff --git a/src/Ophis/Passes.py b/src/Ophis/Passes.py
index ae8139e..67429b2 100644
--- a/src/Ophis/Passes.py
+++ b/src/Ophis/Passes.py
@@ -366,6 +366,10 @@ class PCTracker(Pass):
     def visitByte(self, node, env):
         env.incPC(len(node.data))
 
+    def visitByteRange(self, node, env):
+        if node.data[1].valid(env):
+            env.incPC(node.data[1].value(env))
+
     def visitWord(self, node, env):
         env.incPC(len(node.data) * 2)
 
@@ -744,6 +748,19 @@ class Assembler(Pass):
         env.incPC(len(node.data))
         self.data += len(node.data)
 
+    def visitByteRange(self, node, env):
+        offset = node.data[0].value(env) + 2
+        length = node.data[1].value(env)
+        if offset < 2:
+            Err.log("Negative offset in .incbin")
+        elif offset + length > len(node.data):
+            Err.log("File too small for .incbin subrange")
+        else:
+            for expr in node.data[offset:(offset + length)]:
+                self.outputbyte(expr, env)
+            env.incPC(length)
+            self.data += length
+
     def visitWord(self, node, env):
         for expr in node.data:
             self.outputword(expr, env)