[API] add GetSyntheticValue

[v-bulle]

Adds GetSyntheticValue to the API on top of GetNonSyntheticValue.

[llvmbot]

@llvm/pr-subscribers-lldb

Author: Vincent Belliard (v-bulle)

Changes

Adds GetSyntheticValue to the API on top of GetNonSyntheticValue.

---

Full diff: https://github.com/llvm/llvm-project/pull/95959.diff

2 Files Affected:

• (modified) lldb/include/lldb/API/SBValue.h (+2)
• (modified) lldb/source/API/SBValue.cpp (+12)

diff --git a/lldb/include/lldb/API/SBValue.h b/lldb/include/lldb/API/SBValue.h
index 65920c76df7a8..bec816fb45184 100644
--- a/lldb/include/lldb/API/SBValue.h
+++ b/lldb/include/lldb/API/SBValue.h
@@ -89,6 +89,8 @@ class LLDB_API SBValue {
 
   lldb::SBValue GetNonSyntheticValue();
 
+  lldb::SBValue GetSyntheticValue();
+
   lldb::DynamicValueType GetPreferDynamicValue();
 
   void SetPreferDynamicValue(lldb::DynamicValueType use_dynamic);
diff --git a/lldb/source/API/SBValue.cpp b/lldb/source/API/SBValue.cpp
index 9d7efba024d11..6b77c0e95cedd 100644
--- a/lldb/source/API/SBValue.cpp
+++ b/lldb/source/API/SBValue.cpp
@@ -761,6 +761,18 @@ lldb::SBValue SBValue::GetNonSyntheticValue() {
   return value_sb;
 }
 
+lldb::SBValue SBValue::GetSyntheticValue() {
+  LLDB_INSTRUMENT_VA(this);
+
+  SBValue value_sb;
+  if (IsValid()) {
+    ValueImplSP proxy_sp(new ValueImpl(m_opaque_sp->GetRootSP(),
+                                       m_opaque_sp->GetUseDynamic(), true));
+    value_sb.SetSP(proxy_sp);
+  }
+  return value_sb;
+}
+
 lldb::DynamicValueType SBValue::GetPreferDynamicValue() {
   LLDB_INSTRUMENT_VA(this);
 

[labath]

We should also add a test case for the new API. You can just take an existing test case with some synthetic values (e.g. TestFormattersSBAPI.py), add some Get(Non)SyntheticValue calls, and make sure they do what they should.

[github-actions]

✅ With the latest revision this PR passed the Python code formatter.

[labath]

Thanks.

[labath]

Suggested change

	)
	)
	self.dbg.GetCategory("CCCSynth").SetEnabled(False)

I'm not sure how lldb chooses the summary providers in case of conflicts, but since that's not what we're testing here, we might as well remove the ambiguity.

[v-bulle]

I don't think we can have a conflict. In this test we have several categories which are enabled one after the other. I just add a new one (with a new name CCCSynth2).

[labath]

You have two categories, but they both have a summary provider for the same type (CCC). (Enabling one category does not automatically disable the other ones.) Right now, lldb seems to pick the one you want, but I don't think we promise that anywhere (in fact, I wouldn't be surprised if this comes down to some (nondeterministic) ordering in some hash container). By disabling the first category, we make sure that lldb always pick the one we want (in case, e.g., the container ordering changes). Alternatively, you could just attach the summary provider to a different type (CCC2?)

[jimingham]

The one thing about this that strikes me as odd is that if you have an SBValue that doesn't have a Synthetic Value, this will return the non-synthetic value. The GetNonSyntheticValue didn't have this problem because there's always a non-synthetic value, so provided the SBValue was valid you could always hand that out.

Would it make more sense to make the ValueImpl that's going to represent the synthetic value then check whether that really is synthetic, and return an empty SBValue if it is not?

[labath]

Would it make more sense to make the ValueImpl that's going to represent the synthetic value then check whether that really is synthetic, and return an empty SBValue if it is not?

Sounds like a good idea to me.

[v-bulle]

The one thing about this that strikes me as odd is that if you have an SBValue that doesn't have a Synthetic Value, this will return the non-synthetic value. The GetNonSyntheticValue didn't have this problem because there's always a non-synthetic value, so provided the SBValue was valid you could always hand that out.

Would it make more sense to make the ValueImpl that's going to represent the synthetic value then check whether that really is synthetic, and return an empty SBValue if it is not?

done

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[jimingham]

LGTM