FIX: Advance to next variable if type_info is missing

[oleid]

This prevents an infinite loop if the type info for the current type is missing. The modified unit test demonstrates the issue.

fixes #18

[oleid]

@DanielT Another option to this fix would be returning the variable without type info, but advancing nonetheless. But then one would need to handle the missing type info at a different place. Not sure what is preferable. That change would look like this:

diff --git a/src/dwarf/iter.rs b/src/dwarf/iter.rs
index 6d0ec36..e914f5e 100644
--- a/src/dwarf/iter.rs
+++ b/src/dwarf/iter.rs
@@ -196,6 +196,7 @@ impl<'a> Iterator for VariablesIterator<'a> {
                     self.type_iter = Some(TypeInfoIter::new(ti));
                 } else {
                     self.type_iter = None;
+                    self.current_var = self.var_iter.next();
                 }
                 Some((varname.to_string(), typeinfo, *address))
             } else {
@@ -288,6 +289,14 @@ fn test_varinfo_iter() {
             typeref: 1,
         },
     );
+    variables.insert(
+        "var_d_wo_type_info".to_string(),
+        VarInfo {
+            address: 4,
+            typeref: 404, // some number with no correspondence in the types hash map
+        },
+    );
+
     let mut types = HashMap::<usize, TypeInfo>::new();
     let mut structmembers: HashMap<String, (TypeInfo, u64)> = HashMap::new();
     structmembers.insert("member_1".to_string(), (TypeInfo::Uint8, 0));
@@ -309,5 +318,5 @@ fn test_varinfo_iter() {
     for item in iter {
         println!("{}", item.0);
     }
-    assert_eq!(VariablesIterator::new(&debugdata).count(), 5);
+    assert_eq!(VariablesIterator::new(&debugdata).count(), 6);
 }

[DanielT]

Thanks for the fix!

I prefer the second approach, where variables without typeinfo are still returned by the iterator. Later on I intend to extend the insert code to handle variables without types. After all we do have addresses for them, so pretending such variables have some simple type (char?) and creating a measurement/characteristic is already useful.

[oleid]

I prefer the second approach, where variables without typeinfo are still returned by the iterator.

Here you are. The PR is updated.

[DanielT]

Thanks!