Stack overflow when returning intermediate object to serialize in_addr about glaze

[stephenberry]

What you are doing looks correct. From debugging quickly it looks like this is an issue with nested lambda calls in Glaze. The glz::meta value is invoking a lambda that invokes the FixedName lambda. This should work, but it appears to be misinterpreting the nested lambda as a nullable. So, yes this is a bug in Glaze. But, it shouldn't be too difficult to fix.

[stephenberry]

Ah, I discovered this is not a Glaze bug, but rather a bug in your code. Your glz::meta<in_addr> returns an Rvalue of FixedName<INET_ADDRSTRLEN>. However, your glaze definition in FixedName only accepts Lvalues. If you change [](FixedName& self) to [](cosnt FixedName& self) or [](FixedName&& self), to either accept a const lvalue or an rvalue then your code will run as desired.

This does bring up the question of whether we can catch these errors at compile time. But, I don't currently see a straightforward solution, because the lambda approach needs to be extremely flexible and support rvalues, lvalues, etc. It can also return nullable types and function pointers, so it's very hard to determine what is appropriate. Eventually we may be able to improve the code to catch these errors, but for now make sure that your functions support rvalues.

[stephenberry]

I added a test case based on your example to the code with a comment about the challenges of detecting this user error. #996

[stephenberry]

Glad you're finding the library useful. Let me know if you run into any more issues.

[kalmjasper]

Thanks it's working perfectly now! Makes sense that you can't cast an rvalue in an lvalue.

For my understanding: How does this not result in a compile time bug but rather in a runtime bug? Passing an rvalue to a function that only accepts an lvalue seems like a compile time thing to catch

[stephenberry]

From my debugging it looks like the lambda is convertible to an lvalue function pointer, so it continually tries to deference and then create a new pointer to the lambda. I think this bug could probably be caught, but it's confusing to debug.