Update deprecated builtins __has_trivial to __is_trivially

[aaronmondal]

The old builtins have been depracated in Clang 16. Since they are also present in GCC it should be fine to just use the new variants.

Fixes Issue #1201.

[aaronmondal]

Well that seems to fail 🤣
@derekmauro I tested this with Clang and GCC via bazel test ... Is there some other command i need to run to test this?

[aaronmondal]

Maybe tests are not actually failing but the CI servers are just unreachable. At least the Details links just give me fusion2’s server IP address could not be found.

[derekmauro]

CI is working correctly, but the results are IP blocked. (There is a way to make the results public, but I haven't had time to figure out how to do that.)

I actually thought cfe27e7 fixed this issue. The remaining uses of __has_trivial should be preprocessor guarded. What platform is demonstrating the issue that this change addresses?

I think the issue with your change is that some of the __is_trivially functions take two arguments. If I recall correctly, the person who committed cfe27e7 said that the simpler fix (the one this change is proposing) had some tricky problems, so we decided to just preprocessor guard the old code.

[aaronmondal]

@derekmauro The issue arose with Upstream Clang/LLVM, but we were using an abseil version before cfe27e7.

I'll see if cfe27e7 already fixes the issue. If so, I'll close this PR.

Regardless, I think it is strange that bazel test ... passes. I would expect obvious errors such as wrong argcounts to immediately be detected by it.

[aaronmondal]

cfe27e7 already fixed the issue. Cool! Closing.

[keith]

There seem to be a few more cases here #1289