Adding pragma llvm_inline_ir.
Improved the error messages.
Append "ret void" when the return type is void
Improved the error message
in case when the string passed as llvm inline ir isn't valid llvm
assembly language.
LLVM 3.2 fix.
Add attribute AlwaysInline inside DtoInlineIRFunction.
Always generate a body for llvm_inline_ir
Also, always make llvm_inline_ir functions linkonce_odr. Because
the body is always generated when a module uses a llvm_inline_ir
function, the fact that the linker removes the function shouldn't
cause problems.
Such uncommented changes to the frontend are potentially a huge
source of confusion during merging new frontend versions
(especially if the piece of code gets changed again before the
next upstream release).
This fixes the std.format unit tests for assertion-enabled
LLVM builds.
What the actual solution for the problem should be depends
entirely on the outcome of the "Nested Unions?" discussion
on dmd-internals.
The previous solution was problematic because the change was language-
visble via the isUnsigned trait. Now, pointers are simply explicitly
checked for in the relevant places. (Note that there might be cases in
the diff where a direct isunsigned() call would have been appropriate –
I only checked for instances where the type clearly cannot be a pointer,
but chose to go the safe route in replicating existing behavior
otherwise).
Fixes DMD testcase 'traits'.
A lot of system specific knowledge is already present in LLVM. This is used to populate several fields in global.params instead of hard coded values in main(). Ensures that the frontend and LLVM have always the same values.
The string representation of the data layout is retrieved from the TargetData class (in main) and passed via global.params.dataLayout to the module. Since the gTargetData is also a global variable it makes no sense to pass this information using another global variable.
Implementing the recursion in DeclarationExp::apply, which seems more
natural, lead to a "cannot interpret" CTFE error in Phobos.
An upstream bug will be raised about this.
Some files in our copy of the DMD 2 source had CRLF line endings.
This was not only inconsistent, but also made merging DMD patches
(where LF is used throughout) unnecessarily painful.
The Windows linker LINK insists on the .obj extension. The following changes are made:
- CMake uses the same extension as the C compiler
- global.obj_ext_alt (aka .obj) is recognized as objectfile extension
- global.obj_ext_alt is used on Windows
- New functions codeGenOptLevel() and verifyModule() to remove code duplication
- Hidden option no-verify renamed to disable-verify and moved to optimizer (like opt tool)
- Removed global.params.noVerify
David Nadlinger