Fix for issue #553.

[redstar]

Check if a static array is accessed with a known index. In this case the bounds check can be omitted.

A bit irritating is the comment:

// static arrays could get static checks for static indices
// but shouldn't since it might be generic code that's never executed

because I do not see where the static bounds and indices might be changed later.

[yebblies]

Maybe it's best to wait until yebblies/dmd@186acf4 gets merged into ldc.

[redstar]

The pull request would make life easier. DMD does not generate bounds checking code for static arrays and I try do to the same. :-)
Any chance that the pull request will be merged soon?

[dnadlinger]

@redstar: Regarding your question about the comment, maybe DMD didn't implement the proper checks for static arrays in the frontend yet. And if int[3] a; a[4] = 1; wasn't considered invalid code (but merely undefined behavior when executed), it would have been a mistake for LDC to error out on that.

[redstar]

@klickverbot Thanks for the explanation. I relaxed the check a bit. If the static indices are found to be outside of the static array, a warning is emitted and a runtime check is generated.

[dnadlinger]

caught/detected

[yebblies]

Any chance that the pull request will be merged soon?

@redstar That pull request has already been merged into dmd, just not ldc.

[redstar]

@yebblies Thanks for the pointer. I think I will create another dot-release of 0.12 therefore I like to have this fixed now. But for the next major release I will for sure use skipboundscheck!