ROL_Ptr.hpp std::is_pointer compile error

[ibaned]

@trilinos/rol
@trilinos/phalanx

Expectations

ROL headers should not cause compile errors when included in applications such as Albany.

Current Behavior

Building Albany gives the following compile error from ROL_Ptr.hpp:

build/TrilinosInstall/include/ROL_Ptr.hpp:71:8: error: partial specialization of 'struct std::is_pointer <Teuchos::RCP<PHX::DataLayout>>' after instantiation of 'struct std::is_pointer Teuchos::RCP<PHX::DataLayout>>' [-fpermissive]

https://my.cdash.org/viewBuildError.php?buildid=1353633

Motivation and Context

Albany cannot be compiled due to this error, including for nightly testing.

Definition of Done

This compile error should be fixed

[ibaned]

The code in question:

namespace std {

template<class T>
struct is_pointer<ROL::Ptr<T>> : public std::true_type { };

}

In general, I think specializing std::is_pointer is bad design. Why not just define ROL::is_pointer? We wouldn't want to break downstream code that assumes normal behavior of std::is_pointer. Note that std::is_pointer isn't even true for std::shared_ptr and others, so this really changes its meaning too much.

[ibaned]

It looks like this has nothing to do with Phalanx, sorry for spamming the Phalanx team.

[dridzal]

I absolutely agree with @ibaned . @gregvw , could you please comment on this? This is somewhat urgent.

[gregvw]

This is vestigial code from initial experimentation that slipped through the cracks and needs to be removed. Pushed fix

[ibaned]

@gregvw Thank you!
I don't see a commit in the Trilinos masterdevelop branch though.
Does ROL live in a separate repository and get snapshotted into Trilinos?

[dridzal]

ROL is in a separate repo, HOWEVER it is not snapshotted, it is merged. I was about to start a test cycle to sync and merge. @ibaned , Is it safe to do so, or have you already pushed a fix in the Trilinos repo? This would cause a merge conflict.

[ibaned]

@dridzal I didn't push anything directly, I just opened PR #2177. If you're going to do a test/sync/merge, then lets just ignore that PR and we should be fine.

[dridzal]

Sounds good. I’ll start the process.

[dridzal]

The fix by @gregvw should be in. Please let us know if this takes care of the Albany issue.

[ibaned]

@dridzal @gregvw thank you, Albany seems to be compiling again