Stephan T. Lavavej comments

Results 354 comments of


                                            Stephan T. Lavavej

<filesystem>: prevent filesystem::path dangerous conversions to/from default code page encoding

`std::filesystem::path` permits construction from a source string and a locale. (This doesn't affect conversions from the native `wchar_t` to `char`, if requested.)

<filesystem>: prevent filesystem::path dangerous conversions to/from default code page encoding

We talked about this, and we'll review a PR to add such a warning, off-by-default, through the `deprecated` attribute mechanism (properly phrased as this is neither an ISO nor a...

Should we require SSE2?

GitHub's happy now. For the MSVC-internal build, I also had to edit `src/vctools/langapi/fmath/fmath.c` to remove two `#if ((FMATH_DEBUG) && (TARGET_X86) && (_M_IX86))` blocks implementing an assertion that was assuming `/arch:IA32`....

Should we require SSE2?

I figured out how to fix this with internal PRs: * [AddressSanitizer-PR-554707](https://devdiv.visualstudio.com/DevDiv/_git/AddressSanitizer/pullrequest/554707) * `prod/be` MSVC-PR-554955

`<algorithm>`: `ranges::minmax` accidentally constructed `remove_cvref_t<iter_reference_t<I>>`

Good catch - we talked about this at the weekly maintainer meeting and agree that this is a bug. This also indicates that we have missing test coverage of ranges...

STL: We should _STD qualify _Ugly function calls to avoid ADL

That example is compelling, although the thought of changing all `_Ugly` function calls fills me with despair (more so than `operator&` and `operator,`). It seems that there are a lot...

STL: We should _STD qualify _Ugly function calls to avoid ADL

In that case, I think I want the LWG to tell us what to do.

STL: We should _STD qualify _Ugly function calls to avoid ADL

We talked about this at the weekly maintainer meeting, and we've decided to qualify (with `_STD`, `_RANGES`, `_CHRONO`, etc.) every `_Ugly` function call, like what we currently do for non-`_Ugly`...

Use [[no_unique_address]] after Clang supports it on Windows

MSVC has implemented `[[msvc::no_unique_address]]` which is unconditionally available. However, it's probably too late to request Clang support and take advantage of it before the C++20 ABI lockdown of Ranges/Format/Chronat happens....

P1222R4 `<flat_set>`

We're starting a multi-collaborator feature branch, `feature/flat_set`, with @Saalvage's implementation, so that it can be refined and test coverage can be added.