<xloctime>: time_get::get can still assert 'istreambuf_iterator is not dereferenceable' when the format is longer than the stream

[Alikhalesi]

Resolves #2848

[ghost]

All CLA requirements met.

[fsb4000]

Thanks again for your hard work. I left some comments about the code style.

[MattStephanson]

If you're interested, this might be a good opportunity to harmonize the loops here in _Getfmt and in get. I believe the latter is closer to what the Standard requires in [locale.time.get.members]/8. _Getfmt has some additional assumptions for performance purposes, but I think the overall logic should be the same, so it might be more clear and maintainable if the two loops are doing comparable tests in the same order.

If you don't want to go that route, though, I still think it would be better to have _First == _Last as a standalone test at the beginning of the loop, for performance reasons. IOW, if (*_Fmtfirst == '%') would become an else if following the _First == _Last test. Otherwise, we could make several calls to do_get even after the stream is eof.

[Alikhalesi]

If you're interested, this might be a good opportunity to harmonize the loops here in _Getfmt and in get. I believe the latter is closer to what the Standard requires in [locale.time.get.members]/8. _Getfmt has some additional assumptions for performance purposes, but I think the overall logic should be the same, so it might be more clear and maintainable if the two loops are doing comparable tests in the same order.

If you don't want to go that route, though, I still think it would be better to have _First == _Last as a standalone test at the beginning of the loop, for performance reasons. IOW, if (*_Fmtfirst == '%') would become an else if following the _First == _Last test. Otherwise, we could make several calls to do_get even after the stream is eof.

Would you suggest a good comment for this standalone test?

[MattStephanson]

Would you suggest a good comment for this standalone test?

This isn't a test as in "unit test", so no comment needed, in my opinion.

[MattStephanson]

LGTM!

[StephanTLavavej]

Looks good to me too. I agree that testing _First == _Last at the beginning of each loop iteration follows [locale.time.get.members]/8 - in particular, that this happens before the whitespace-skipping logic.

[strega-nil-ms]

This seems clearly correct according to the standardese.

[StephanTLavavej]

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

[StephanTLavavej]

Thanks for fixing this runtime correctness bug, and congratulations on your second microsoft/STL commit! 🎉 😻 🚀

This will ship in VS 2022 17.4 Preview 2.