Make ExceptionFrames fields private

[jonas-schievink]

Closes #234

I can also add the unsafe setters, but they don't have any use right now. (added them in order to not regress available operations on ExceptionFrame)

[rust-highfive]

r? @korken89

(rust_highfive has picked a reviewer for you, use r? to override)

[korken89]

LGTM, does anyone else want to chime in?

[Disasm]

Is there any sense in modifying only some of the full set of registers? If you implement an RTOS, you may want to save/restore all of the registers.

[jonas-schievink]

Is there any sense in modifying only some of the full set of registers? If you implement an RTOS, you may want to save/restore all of the registers.

Yes, I've had to set only pc in RTOS code, to jump into userspace once the interrupt returns. Also, for SVCall, you generally want to read the registers that pass arguments to the syscall, and then write back the result (which can end up in the same set of registers, or only r0). In my case these registers were r0-3.

Saving / Restoring all registers isn't a bad point though, since that's needed for context switching. Maybe we can provide wrappers around pointer casts / transmute / ptr::{read, write} that can be used to deal with entire ExceptionFrames at a time? I'm open to suggestions here.

(this reminds me that none of these functions are #[inline], I should definitely fix that before this is merged)

[japaric]

This looks good to me.

If we want to support the RTOS case of modifying stacked registers in this minor release we can add a ExceptionFrameMut struct with this API instead of modifying the existing ExceptionFrame. Whether we want to do that or not depends on how far the next minor version is and how much need there is for ExceptionFrameMut right now.

[thejpster]

bors r+

[jonas-schievink]

bors r-

inline attributes are still missing

[bors]

Canceled

[korken89]

bors r+

[jonas-schievink]

bors r=korken89

[bors]

Build succeeded

• continuous-integration/travis-ci/push