[Lldb-commits] [lldb] [lldb] Add Function::GetAddress and redirect some uses (PR #115836)

[Pavel Labath via lldb-commits]

================
@@ -407,6 +406,15 @@ CompileUnit *Function::GetCompileUnit() { return m_comp_unit; }
 
 const CompileUnit *Function::GetCompileUnit() const { return m_comp_unit; }
 
+Address Function::GetAddress() const {
+  if (m_ranges.empty())
+    return Address();
+  // We're using a (DWARF-like) convention where the base address of the first
+  // interval denotes the entry point of the function. If that turns out to be
+  // insufficient, we can introduce a separate "entry point address" field.
+  return m_ranges[0].GetBaseAddress();
----------------
labath wrote:

There's no guarantee that the function's entry point will be its lowest numbered address. DWARF even allows you (via `DW_AT_entry_pc`) to have the entry pc point to the middle of a range. I don't think any compiler emits that, and lldb doesn't support that right now, but if we wanted to do it, then we'd need to have a separate `m_address` field here independent of any range (that's what I'm alluding to in that comment).

The "base address for relocation of blocks and variables" use case is totally under our control, so we could choose that to be based on the lowest address of the function, but that would mean we'd have to have two accessors (and then be careful about which one is used where). I think it's better to use the same address for both -- and accept the fact that some of the block/variable offsets may be negative. (Having written that, I realize this also means that we'd need to change the block/variable offsets to signed representations. Right now they're unsigned, which I might sort of make this patch a regression for those kinds of functions, but given that discontinuous functions are barely supported anyway, I'd like to do that separately.) This also fits in nicely with DWARF which says that "The base address of the scope for any of the debugging information entries listed
above is given by either the DW_AT_low_pc attribute or the first address in the first range entry in the list of ranges given by the DW_AT_ranges attribute." -- it doesn't care whether the base address is numerically lowest.

Let me know what you think.

https://github.com/llvm/llvm-project/pull/115836