<bits/move.h> is a part of <utility>.

[bungeman]

<bits/move.h> provides std::forward, std::move, and std::swap. These are defined to come from <utility> and not <algorithm>.

In c++11 and later std::swap comes from <utility>, previously it came from <algorithm>. Fortunately, <bits/move.h> is a c++11 only header, so if any symbols are being seen in it, that's where they should come from.

This was reported in #486.

[kimgr]

Thanks! Any chance for a testcase?

Also, we have freestanding mappings in gcc.*.imp, could you make sure they're updated similarly?

[bungeman]

Updated the *.imp, add the test from #486. I'm doing this between other things, so I just did all this in the github webui, resulting in all these one file commits. I think you can just "Squash and merge" and clean up the commit message. If not, let me know and I'll see if there's a good way to squash all this.

[kimgr]

Neat, thanks! I'll take another look, test on my freebsd system and clean it up if I find anything.

[bungeman]

I made it back to a real machine so I squashed this merge request myself. Makes it easier to look at.

[kimgr]

Thank you! I just gave this a whirl on FreeBSD. Of course the test fails :-/. The system C++ library on BSD is libc++ and all the builtin mappings are for libstdc++, so we don't have mappings and <type_traits> is suggested for std::move.

I'm not sure how I feel about this. On the one hand test coverage is nice, but on the other hand non-portable tests are a pain. We don't have good test infrastructure to detect these things either.

We actively try to keep new test cases independent of the environment by creating stub headers for constructs in the standard library, when we run into issues. But the mappings are naturally environment-specific.

Sigh. I'll rework your commit and remove the test (sorry about the wasted effort).

[kimgr]

Did that and merged in 314fc2d. Thanks for the patch!