Merge remote-tracking branch 'origin/topic/robin/gh-1842-offset'

* origin/topic/robin/gh-1842-offset:
  Fix when input redirection becomes visible.

CHANGES

@@ -1,3 +1,12 @@
+1.12.0-dev.116 | 2024-09-30 13:35:49 +0200
+
+  * GH-1842: Fix when input redirection becomes visible. (Robin Sommer, Corelight)
+
+    With `&parse-at/from` we were updating the internal state on our
+    current position immediately, meaning they were visible already when
+    evaluating other attributes on the same field afterwards, which is
+    unexpected.
+
 1.12.0-dev.114 | 2024-09-30 10:24:42 +0200
 
   * GH-1844: Fix nested look-ahead parsing. (Robin Sommer, Corelight)

VERSION

@@ -1 +1 @@
-1.12.0-dev.114
+1.12.0-dev.116

doc/autogen/types/unit.rst

@@ -57,14 +57,13 @@
 
     Returns the offset of the current location in the input stream
     relative to the unit's start. If executed from inside a field hook,
-    the offset will represent the first byte that the field has been
-    parsed from.
+    the offset will represent the beginning of the field, not the end.
 
 .. spicy:method:: unit::position unit position False iterator<stream> ()
 
     Returns an iterator to the current position in the unit's input
     stream. If executed from inside a field hook, the position will
-    represent the first byte that the field has been parsed from.
+    represent the beginning of the field, not the end.
 
 .. spicy:method:: unit::set_input unit set_input False void (i: iterator<stream>)
 

spicy/toolchain/src/ast/operators/unit.cc

@@ -191,7 +191,7 @@ class Offset : public hilti::BuiltInMemberCall {
             .doc = R"(
 Returns the offset of the current location in the input stream relative to the
 unit's start. If executed from inside a field hook, the offset will represent
-the first byte that the field has been parsed from.
+the beginning of the field, not the end.
 )",
         };
     }
@@ -212,8 +212,8 @@ class Position : public hilti::BuiltInMemberCall {
             .ns = "unit",
             .doc = R"(
 Returns an iterator to the current position in the unit's input stream. If
-executed from inside a field hook, the position will represent the first byte
-that the field has been parsed from.
+executed from inside a field hook, the position will represent the beginning of
+the field, not the end.
 )",
         };
     }

spicy/toolchain/src/compiler/codegen/parser-builder.cc

@@ -548,6 +548,7 @@ struct ProductionVisitor : public production::Visitor {
             auto etype = c->parseType()->type()->elementType();
             auto container_element = builder()->addTmp("elem", etype);
             pushDestination(container_element);
+            pb->saveParsePosition(); // need to update position for container elements in case input is redirected
         }
 
         else if ( ! meta.isFieldProduction() )
@@ -1430,7 +1431,6 @@ struct ProductionVisitor : public production::Visitor {
         pstate.cur = builder()->addTmp("parse_cur", builder()->typeStreamView(), builder()->deref(tmp));
         pstate.ncur = {};
         pushState(std::move(pstate));
-        pb->saveParsePosition();
     }
 
     // Redirects input to be read from given stream position next.
@@ -1446,7 +1446,6 @@ struct ProductionVisitor : public production::Visitor {
         pstate.cur = builder()->addTmp("parse_cur", cur);
         pstate.ncur = {};
         pushState(std::move(pstate));
-        pb->saveParsePosition();
     }
 
     // Start sync and trial mode.

tests/Baseline/spicy.types.unit.size-with-offset/output

@@ -0,0 +1,5 @@
+### BTest baseline data generated by btest-diff. Do not edit. Use "btest -U/-u" to update. Requires BTest >= 0.63.
+x, 3
+x, \x01\x02\x03
+z, 4
+z, 1234

tests/spicy/types/unit/offset-with-parse.spicy

@@ -0,0 +1,18 @@
+# @TEST-EXEC: printf 'a' | spicy-driver -d %INPUT
+#
+# @TEST-DOC:
+
+module Test;
+
+public type X = unit {
+    a: bytes &size=1;
+
+    y1: Y &parse-at=self.input();
+    y2: Y &parse-from=self.a; # Behaves identical.
+};
+
+type Y = unit {
+    a: bytes &size=1 {
+        assert self.offset() == 0 : "First element should always be at offset 0";
+    }
+};

tests/spicy/types/unit/size-with-offset.spicy

@@ -0,0 +1,30 @@
+# @TEST-EXEC: ${SCRIPTS}/printf '\x01\x02\x03\x04\0x05' | spicy-driver %INPUT >output
+# @TEST-EXEC: btest-diff output
+#
+# @TEST-EXED: Check that changing the input with `&parse-at` isn't visible to other attributes on same field; regression test for #1842.
+
+module Test;
+
+import spicy;
+
+public type X = unit {
+    a: uint8;
+    b: uint8;
+    c: uint8;
+
+    x: bytes &size=self.offset()  # yields 3
+             &parse-at=self.input() {
+        print "x", self.offset(); # yields still 3
+        print "x", $$;
+    }
+
+    y: uint8;
+
+    z: bytes &size=self.offset()  # yields 4
+             &parse-from=b"12345" {
+        print "z", self.offset(); # yields still 4
+        print "z", $$;
+    }
+
+    d: uint8;
+};