On Tue, Apr 14, 2020 at 10:38:47PM +0200, Simon Goldschmidt wrote:
Masahiro Yamada masahiroy@kernel.org schrieb am Di., 14. Apr. 2020, 17:34:
Hi Simon,
On Tue, Apr 14, 2020 at 11:41 PM Simon Glass sjg@chromium.org wrote:
Hi Masahiro,
On Mon, 13 Apr 2020 at 11:10, Masahiro Yamada masahiroy@kernel.org
wrote:
On Fri, Apr 10, 2020 at 5:18 AM Simon Glass sjg@chromium.org wrote:
Some places use __ASSEMBLER__ instead which does not work since the Makefile does not define it. Fix them.
In my understanding, __ASSEMBLER__ is passed by the compiler while building *.S files.
On which compiler didn't this work for you?
Signed-off-by: Simon Glass sjg@chromium.org
Changes in v2:
- Add new patch to fix occurances of __ASSEMBLER__
I think I hit this with device tree, but I would need to go back and check. Anyway I think we should be consistent.
DT files are pre-processed with '-x assembler-with-cpp'.
So, __ASSEMBLER__ is pre-defined by the compiler.
I have just seen this thread now. It touches socfpga headers but we're not in CC (possibly the maintainers file needs updating).
Anyway, why do we have 2 defines for assembler? If one is predefined by the compiler, why don't we use that? Can you explain that for someone not too deeply involved?
Generally and broadly speaking, the assembler will set __ASSEMBLER__. That said, the Linux kernel has been setting __ASSEMBLY__ for practically forever and we've copied that behavior also practically for forever. In both trees there are a handful of __ASSEMBLER__ tests and a vast majority of __ASSEMBLY__ tests. It doesn't normally matter which is used as the only non-compiler in the Linux kernel that might care is handled as shown above. Under the assumption that the feature Yamada-san pointed out in binman is used, it should also define __ASSEMBLER__ I believe, to match other tooling. Thanks!