Add note for array with empty struct or union

[kito-cheng]

Fix #358

The key point of this issue is C and C++ having different type size for array of empty struct or union, size of empty struct or union is 0 for C, and 1 for C++, and the size of array is type size multiplied by the length of the array.

Also checked AArch64 code gen, they also generated different code between C and C++, and ABI isn't explicitly describe this but has describe the size of array is type * length.

[asb]

Thanks for looking at this Kito. Quick high level comment: I think the case of a zero-length array of empty structs/unions should also be mentioned. Perhaps it's worth being more explicit about the case of a 1-sized array of an empty struct/union too?

[kito-cheng]

Changes:

• Add example to show the case struct { struct {} e[1]; float f; };
• Add sentence to mention zero-length array will ignored during struct flatten.

[kito-cheng]

@nick-knight thanks for your comment and suggestion :)

[kito-cheng]

@asb @jrtc27 @luismarques did you mind take a review for that? I would like to have your review before I merge that :)

[luismarques]

Sorry, I shouldn't have approved this. What is there LGTM but it seems it doesn't explicitly cover all cases that @asb was asking about. Can you be more explicit about both the [0] and the [1] cases?

[kito-cheng]

@luismarques added more description for zero-length array of empty structs/union, thank for the review :)

[kito-cheng]

Changes:

• Explicitly say the field with array of empty struct type will be ignored or not in the example.

[asb]

@kito-cheng the last round of grammatical suggestions from @nick-knight are marked as resolved but the changes don't seem to be reflected in the current version of the patch. It looks like they might have got lot along the way?

I think I may have some additional minor grammatical or phrasing tweak suggestions after that, but it might be inefficient to comment if the version of the patch in this PR is an older version.

[kito-cheng]

@asb thanks for catch that...I must did something wrong during update this PR, updated now :)

[asb]

LGTM modulo the minor wording fixes suggested in line. Thanks Kito!

[kito-cheng]

Thanks everyone :)