Update numeric.d

[9il]

Modify private oppositeSigns.
Reason: findRoot!(double, real) failed to compile.

[9il]

Error (without fix)

/usr/include/dmd/phobos/std/numeric.d(837): Error: template std.numeric.oppositeSigns does not match any function template declaration. Candidates are:
/usr/include/dmd/phobos/std/numeric.d(682):        std.numeric.oppositeSigns(T)(T a, T b)
/usr/include/dmd/phobos/std/numeric.d(837): Error: template std.numeric.oppositeSigns(T)(T a, T b) cannot deduce template function from argument types !()(double, real)
/usr/include/dmd/phobos/std/numeric.d(715): Error: template instance std.numeric.findRoot!(double, real) error instantiating
source/app.d(307):        instantiated from here: findRoot!(double, real)
source/app.d(307): Error: template instance std.numeric.findRoot!(double, real) error instantiating

[Poita]

LGTM

Edit: there should be a unit test for the bug to protect against regressions.

[ghost]

The usual convention is T1/T2/etc.., not Ta/Tb.

[9il]

@Poita, unittest added.
@AndrejMitrovic , T1/T2 fixed.

[9il]

One test failed after T1 and T2 fix:
Internal error: backend/cg87.c 925.
Probably it is unrelated bug.

[ghost]

Could you file the ICE? Thanks.

[9il]

https://d.puremagic.com/issues/show_bug.cgi?id=12057

[monarchdodra]

The usual convention is T1/T2/etc.., not Ta/Tb.

Yeah, that or T and U (and V). The "issue" with T1/T2, is that if you change them to T..., then T1 becomes T[0], and T2 becomes T[1]. Not really a huge issue here, but I have encountered this once or twice when "upgrading" code before. Kind of off topic, but just wanted to throw that in there. Something to think about.

[ghost]

Btw where does T/U/V come from?

[ghost]

Something to think about.

Hmm yep. It's funny how we're used to 0-indexing but we often use T1/T2 rather than T0/T1.

[monarchdodra]

Btw where does T/U/V come from?

Simply the alphabet I guess. Kind of like we start with x, and then we have y and z for dimensions. For Meta programming, we started with T, and the next logical letter is simply U.

[monarchdodra]

Hmm yep. It's funny how we're used to 0-indexing but we often use T1/T2 rather than T0/T1.

In C++, migrating from pair to tuple is a bit painful for at first because of that too: The member names are literally first and second!

[9il]

I will create new PR because I delete my clone of Phobos.