Remove `-Werror=missing-braces`

[fwyzard]

-Werror=missing-braces prevents CTAD:

template <std::integral... Args>
void test(Args... args) {
    auto x = std::array{static_cast<int>(args)...};
}

int main() {
    test(0, 1, 2u, 3u);
}

Once the second level of { ... } is added, it is necessary to specify the full array type:

template <std::integral... Args>
void test(Args... args) {
    auto x = std::array<int, sizeof...(Args)>{{static_cast<int>(args)...}};
}

int main() {
    test(0, 1, 2u, 3u);
}

Since c++20 seems to support this, I would argue to remove the -Werror flag.

[cmsbuild]

cms-bot internal usage

[cmsbuild]

A new Issue was created by @fwyzard.

@Dr15Jones, @ftenchini, @makortel, @mandrenguyen, @sextonkennedy, @smuzaffar can you please review it and eventually sign/assign? Thanks.

cms-bot commands are listed here

[smuzaffar]

assign core

[cmsbuild]

New categories assigned: core

@Dr15Jones,@makortel,@smuzaffar you have been requested to review this Pull request/Issue and eventually sign? Thanks

[smuzaffar]

cms-sw/cmsdist#10141 removes this flags and cms-sw/cmsdist#10141 (comment) build results look good

[smuzaffar]

@makortel , any objections on getting this in for 16.0.X ? Tests for cms-sw/cmsdist#10141 look good

[makortel]

I'm a little bit concerned of the tradeoff of not being able to flag cases that would be confusing without the additional braces like

int a[2][2] = { 0, 1, 2, 3 };
// vs
int b[2][2] = { { 0, 1 }, { 2, 3 } };

(from GCC manual. I don't know though how to estimate if this error has been useful in the past, and if it has, how useful. In principle such confusing code should be noticed in code review, but from the past experience I wouldn't expect many such cases to be caught, unfortunately.

Written that I agree with the second form in the issue description being annoying. For the specific purpose of std::array, I found C++20 introduced std::to_array() that could be used along

template <std::integral... Args>
void test(Args... args) {
    auto x = std::to_array({static_cast<int>(args)...});
}

int main() {
    test(0, 1, 2u, 3u);
}

(not exactly ideal either, but would avoid the need to specify the value type and size for the array)

So I'm not outright objecting, but I am a bit worried if not having that error would bite us in the future (e.g. in a form things being initialized incorrectly because of the initialization expression looking confusing).

[makortel]

Let's discuss this in the core software meeting on Tuesday

[makortel]

The decision was to proceed with removal of this error flag.