Remove packed attribute for uint32_t (new GCC-8 warning)

[prusnak]

New GCC-8 complains about using packed attribute for uint32_t. This patch removes this extraneous usage.

[dpgeorge]

Thanks for the patch. We need to be careful with things like this because the HAL can be picky when it comes to these kinds of things (see eg 1fe30d1).

I think the __packed attribute here is intended to mean that the uint32_t* pointer may be unaligned, and that's important to keep, so I'd be hesitant about removing the attribute.

With arm-none-eabi-gcc version 8.2.0 I don't get any warnings. Which version do you use?

[prusnak]

arm-none-eabi-gcc (GNU Tools for Arm Embedded Processors 8-2018-q4-major) 8.2.1 20181213 (release) [gcc-8-branch revision 267074]

downloaded from: https://developer.arm.com/open-source/gnu-toolchain/gnu-rm/downloads

[dpgeorge]

Doing some research, it seems that ARM CC accepts the __packed attribute and takes it to mean "this pointer may be unaligned", see http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0472m/chr1359124968737.html

Then for gcc there is a compatibility macro in the HAL which defines __packed to attribute((packed)). But gcc doesn't support this usage, since to it packed means don't add padding in a struct. So I guess that gcc 8.2.1 is now warning about this bad usage.

https://answers.launchpad.net/gcc-arm-embedded/+question/452880 contains some more info about gcc in such a case, and the last post on that page provides a way to get it working in gcc by defining a packet struct with a single member (in this case it would be a uint32_t).

---

I'm assuming that this packed qualifier never did anything on gcc (although that remains to be checked) and so I think the best way to handle it is to #define __packed to nothing for the case of gcc.

[prusnak]

It seems the __packed is indeed used only for uint32_t in the whole stm32lib codebase, so it might make sense to undefine it for GCC in this lib.

[dpgeorge]

It seems the __packed is indeed used only for uint32_t in the whole stm32lib codebase, so it might make sense to undefine it for GCC in this lib.

How about just undefining it in these ll_usb.c files? I.e. at the top of these .c files: #if GCC, #undef __packed. That is the smallest change (so least unforeseen consequences) that solves the problem.

[prusnak]

I agree.

[prusnak]

Force pushed to my branch

[dpgeorge]

Sorry to be a bother, but I just realised this will need to be split into 4 separate commits (which can be all put in this same PR), one for each of the MCU variants. That makes it easier to maintain in the future, when the Cube HAL is updated. (Existing commits do this.)

It'll also be good to add a C comment to this change, something like: In this file __packed is used to signify an unaligned pointer, which GCC doesn't support, so disable it.

Thanks!

[prusnak]

Updated the PR

[prusnak]

I realised this should not go to the vendor branch but rather to the latest active work-* branch. Superseded by #7