mbed fault handler: fix mbed_fault_context parameter

[andrewc-arm]

Description

Why required

Without this fix, when void mbed_fault_handler(uint32_t fault_type, void *mbed_fault_context_in) is called, mbed_fault_context_in is in **mbed_fault_context_t type which is unnecessarily doubly pointered. Also, it's unintuitive, most users will expect to simply static cast the void pointer argument to *mbed_fault_context_t instead of **mbed_fault_context_t.

What's the change

The except.S Fault_Handler assembly code will prepare mbed_fault_context_in argument as *mbed_fault_context_t type instead of **mbed_fault_context_t type.

Pull request type

[x] Fix
[ ] Refactor
[ ] Target update
[ ] Functionality change
[ ] Docs update
[ ] Test update
[ ] Breaking change

Reviewers

Release Notes

[0xc0170]

Can you rework the commit message, shall be something like: mbed fault handler: fix mbed_fault_context parameter and second paragraph containing info about the fix - what is it fixing and why

[kjbracey]

If you're going to change this, it's a behaviour change, so you should at least change the parameter to be mbed_fault_context_t * to show what it actually is. (Maybe const, if the handler isn't supposed to modify it).

This isn't public API anyway, I believe, and it doesn't currently use that parameter, so I'd be inclined to just stop passing that parameter at all - save an instruction.

While here, remove the unnecessary cast at

mbed-os/platform/source/TARGET_CORTEX_M/mbed_fault_handler.c

Line 39 in ff4dac9

mbed_fault_context_t *const mbed_fault_context = (mbed_fault_context_t *) &fault_context;

Is there a reason this change is happening?

[andrewc-arm]

Is there a reason this change is happening?

I was making stack dump feature for my customers and found that parameter was returning misleading information. It took me quite sometime to figure out the right values.

I agree with const mbed_fault_context_t *. It would really help users to understand what it is instead searching through the code/assembly. I will change them. For the mbed_fault_handler override, they can make use of the argument well.

[ciarmcom]

@andrewc-arm, thank you for your changes.
@ARMmbed/mbed-os-core @ARMmbed/mbed-os-maintainers please review.

[kjbracey]

Looks fine to me, but it is an API change, so not sure it would be appropriate for a patch release. But I don't feel that strongly.

[0xc0170]

@andrewc-arm Lets take it to 5.14. If this is API change, should be "functional change" in PR type above and contain "release notes" details as well.

[andrewc-arm]

Sure. Thanks for making better software.

[0xc0170]

CI started

[mbed-ci]

Test run: SUCCESS

Summary: 11 of 11 test jobs passed
Build number : 1
Build artifacts