bc: use __has_attribute to test attribute presence

[rlibby]

Summary

Suppress empty declaration error on gcc versions 6 and lower, which
don't have the fallthrough attribute.

Test Plan

env MAKEOBJDIRPREFIX=/usr/obj/gcc6 CROSS_TOOLCHAIN=amd64-gcc make buildworld
env MAKEOBJDIRPREFIX=/usr/obj/clang make buildworld

[rlibby]

This has an upstream but I've submitted the patch here because that seems to be what the upstream author has requested. Please let me know if we should engage upstream first.

https://git.yzena.com/gavin/bc#user-content-other-projects

[se]

I have updated the math/gh-bc port to version 3.2.4, which includes a slightly modified version of this change.
The author wanted to avoid problems with other compilers than gcc and clang, which might not support the __has_attribute(()) mechanism, and therefore has added further conditions to your patch.

Warner Losh has asked to avoid vendor imports until after the conversion to git, else I'd resolve this issue with a vendor import and MFV,
A patch to contrib/bc could be committed instead, if you consider the issue urgent enough to not allow waiting for 2 weeks.

If you consider committing a fix that urgent, then I'd ask you to use the following patch, in order to not cause a conflict in a later MFV (I cannot update the diff in this review to this version and did not want to create another review with this patch):

Index: contrib/bc/include/status.h
===================================================================
--- contrib/bc/include/status.h	(revision 368505)
+++ contrib/bc/include/status.h	(working copy)
@@ -166,7 +166,11 @@
 #endif // __STDC_VERSION__
 
 #if defined(__clang__) || defined(__GNUC__)
+#if defined(__has_attribute) && __has_attribute(fallthrough)
 #define BC_FALLTHROUGH __attribute__((fallthrough));
+#else // defined(__has_attribute) && __has_attribute(fallthrough)
+#define BC_FALLTHROUGH
+#endif // defined(__has_attribute) && __has_attribute(fallthrough)
 #else // defined(__clang__) || defined(__GNUC__)
 #define BC_FALLTHROUGH
 #endif //defined(__clang__) || defined(__GNUC__)

If you think that a commit can wait 2 weeks, then I'll import a new vendor version after the conversion of the src repository to git ...

[rlibby]

Yeah I think waiting is probably fine. This is one of a few things breaking gcc builds. Some ZFS issues are also causing trouble and I guess those may have the same hiccup with vendor imports.

Thanks.

[rlibby]

@se Now that vendor workflows are open for git, can we get upstream's fix in? Or do you prefer I just do a direct commit of upstream's fix as discussed above?

[se]

In D27562#620824, @rlibby wrote:

@se Now that vendor workflows are open for git, can we get upstream's fix in? Or do you prefer I just do a direct commit of upstream's fix as discussed above?

I plan to do a vendor import and MFV, but have to understand and test the procedure first.
Please allow for 24 hours to perform the import (its past midnight, already, here ...)

[rlibby]

Yep, just a friendly ping. I'm eager to finish the gcc work, but it is definitely not urgent. Thank you!

[rlibby]

Resolved by @se's import of bc 3.2.4.

https://cgit.freebsd.org/src/commit/?id=e458944cf9deec51d03ec751050a58ddf43e796f