Document known issues, minor tweak to tests/p_05.p4

ISSUES.md can track all known issues, starting with the
fact that parse_ctrl.state should be a match parameter
in match.p4t tables, as a ternary match since it won't
be needed all the time.

ISSUES.md

+# Known Issues
+
+## Unsupported: Headers With Multiple Potential Offsets
+
+HyPer4 assumes that every potential parsed representation header
+has just one possible offset, which is not true of P4 programs
+in general.
+
+In hp4.p4, the parse control state (parse_ctrl.state) is not
+a match parameter in the virtual ingress pipeline.
+
+Adding it would require additional development in p4_hp4 and
+hp4controller / ConViDa that is not critical to completing
+support for VIBRANT.  We would need to make the compiler and
+interpreter support the additional rules required for each
+terminal parse control state in which affected headers end
+up at different offsets.  When it is time to implement this,
+we can be smart by making the match parameter in hp4.p4
+ternary, and leaving it turned off for those matches where
+the parse control state does not matter (because the headers
+being matched are always at the same offset despite other
+headers having multiple potential offsets).
+
+For now, we can implement a check in p4_hp4 that detects
+headers with multiple potential offsets, and terminate
+with an error.

tests/p_05.p4

@@ -84,6 +84,9 @@ action _no_op() {
 }
 
 table t {
+  reads {
+    A.f1 : exact;
+  }
   actions {
     _no_op;
   }