[clang] No way to recognize lambda expression in DWARF

[zhihaoy]

GCC's practice is to mark the operator() DW_AT_artificial. It does the job and makes perfect sense (i.e., it's a synthesized operator() that the user did not declare).

(Older GCC also names the unnamed closing class type, but newer GCC no longer does it, and tools have adapted since then).

[llvmbot]

@llvm/issue-subscribers-clang-codegen

Author: None (zhihaoy)

GCC's practice is to mark the `operator()` `DW_AT_artificial`. It does the job and makes perfect sense (i.e., it's a synthesized `operator()` that the user did not declare).

(Older GCC also names the unnamed closing class type, but newer GCC no longer does it, and tools have adapted since then).

[llvmbot]

@llvm/issue-subscribers-debuginfo

Author: None (zhihaoy)

GCC's practice is to mark the `operator()` `DW_AT_artificial`. It does the job and makes perfect sense (i.e., it's a synthesized `operator()` that the user did not declare).

(Older GCC also names the unnamed closing class type, but newer GCC no longer does it, and tools have adapted since then).

[dwblaikie]

What special behaviour follows from the op() being artificial?

[zhihaoy]

What special behaviour follows from the op() being artificial?

This is only to help with tooling. For example, I can hide the operator() and display something terse and sensible. But given the status quo, I can't distinguish it from an unnamed struct with operator().

If you were asking if DWARF associates the artificial flag with any implications to other DWARF data, I think not. I read

2.11 Artificial Entries
Any debugging information entry representing the declaration of an object or type artificially generated by a compiler and not explicitly declared by the source program may have a DW_AT_artificial attribute, which is a flag.

It did not explicitly say "function" here, but in 5.7.8 Member Function Entries explained as follows

An artificial member function (that is, a compiler-generated copy that does not appear in the source) does not have a DW_AT_defaulted attribute.

, which is a widely accepted practice,

And in Appendix A

5. The DW_AT_artificial attribute can be used with any declarative debugging information entry. [...]

[dwblaikie]

Yeah, I do'nt really have a problem with adding DW_AT_artificial to the op() of a lambda class - though I'm not sure how much that'd be good for users of the consumers using this information. As a debugger user, honestly, I'd probably prefer to see the type rendered like other types... but as a compiler engineer - yeah, add the information (that the op() is artificial) and let consumers decide what's best for their users.