Upgrade frontend & libs to pre-DMD-monorepo state #4141
Conversation
|
Apologies for the merge conflict, I've got a couple of changes I need to make to druntime, namely making the intrinsic names for pointers not include the type for LLVM 15 (well technically opaque pointers, which in theory we could enable for LLVM14 or disable for 15) e.g. -pragma(LDC_intrinsic, "llvm.memcpy.p0i8.p0i8.i#")
+pragma(LDC_intrinsic, LLVM_version >= 1500 ? "llvm.memcpy.p0.p0.i#" : "llvm.memcpy.p0i8.p0i8.i#")
void llvm_memcpy(T)(void* dst, const(void)* src, T len, bool volatile_ = false)
if (__traits(isIntegral, T));ditto for |
Using its `foreachVar()` method to visit the nested vars, as done in the DMD glue layer.
Conflicts: gen/toir.cpp runtime/druntime
|
The undefined |
This matters for MSVC, which doesn't ignore the outer `const` for mangling.
…ring length Fixes dmd-testsuite's compilable/test23066.c.
Conflicts: runtime/druntime
…members
Fixes assertions hit by dmd-testsuite's runnable/{test23010,variadic}.d,
caused by a different tuples handling in the frontend.
| @@ -31,7 +31,7 @@ void func() | |||
| int[] quux; | |||
| //CHECK: dcompute.d([[@LINE+1]]): Error: can only call functions from other `@compute` modules in `@compute` code | |||
| quux.length = 1; | |||
| //CHECK: dcompute.d([[@LINE+1]]): Error: cannot use operator `~=` in `@compute` code | |||
| //CHECK: dcompute.d([[@LINE+1]]): Error: can only call functions from other `@compute` modules in `@compute` code | |||
There was a problem hiding this comment.
this looks like a regression to me. @thewilsonator can you help here to keep the better old error message?
There was a problem hiding this comment.
These are caused by new frontend lowerings (the templated druntime 'hooks').
There was a problem hiding this comment.
Perhaps we can hardcode the recognition of names of the hooks, and give a bit nicer feedback to the user. It will not be easy to understand that ~= leads to a "call to a function from another non-@compute module".
There was a problem hiding this comment.
I'm fine with making an open issue for it, shouldn't block the monorepo work.
No description provided.