Fixes: https://github.com/dlang/dmd/issues/21183
The previous MR put the macro in a `#if linux` which meant it
didn't actually solve the problem of being unable to
`#include <math.h>` on macos. So put it in a better spot.
Also enable the test that includes that header for macos so that
it stays solved.
* bump VERSION to v2.110.0
* purge changelog
* bump VERSION to v2.111.0-beta.1
* Accept __rvalue attribute on ref functions; which will force the result to be treated as __rvalue. (#20946)
This is essential to implement `move`, `forward`, etc.
* memoryerror.d: Fix AnySupported version condition (#20983)
* Fix#20982 - wrong line number in iasmgcc (#20993)
* Move genCfunc to cxxfrontend (#20992)
* druntime: Fix compilation of rt.cover on Android (#21015)
* Expose SourceLoc to C++ interface (#20980)
* [stable] C++ header fixes for declaration, expression, and typinf (#21016)
Seen either from compilation errors or missing symbols at link time.
* C++ headers: Add 3 Declaration bitfield setters/getters required by LDC
* druntime: Add module declaration to rt.invariant, to prevent conflicts with user-provided invariant.d (#21017)
* Fix#21020 - Indexing a *cast* AA yields no lvalue anymore (#21029)
* Add C++23 to CppStdRevision enum (#21043)
* Improve UFCS/property error message (#21046)
---------
Co-authored-by: Manu Evans <turkeyman@gmail.com>
Co-authored-by: Martin Kinkelin <kinke@users.noreply.github.com>
Co-authored-by: Iain Buclaw <ibuclaw@gdcproject.org>
Co-authored-by: Martin Kinkelin <noone@nowhere.com>
MSVC does not support C11 atomics, but it still provides a "stub"
version of stdatomic.h, which aborts compilation with an #error
directive when included.
To avoid this error, explicitly set the C standard to C11 when invoking
the C preprocessor on Windows, and check for the C11 macro
__STDC_NO_ATOMICS__ before including stdatomic.h.
- let the compiler generate type info for the AA Entry and add it to TypeInfo_AssociativeArray
- strip modifiers for index and next in TypeInfo_AssociativeArray to make it agnostic to changes by the glue layer
threads. The `thread_preSuspend` hook should return true when druntime
has knowledge of a thread. But it's based on `sm_this` (the storage for
`Thread.getThis`) being set. Because the thread lock is not taken to set
`sm_this`, a race exists when a thread is suspended between this
setting, and the adding to the thread list for scanning. Therefore,
`thread_preSuspend` can return true, but `thread_scanAll` will not
include that thread in the list of scannables.
The current check that special cases static arrays of void doesn't take
type qualifiers into account - e.g. void[42] is treated as having
indirections, but const(void[42]) is not. This fixes that.
