Casting to Double of CGFloat.native for math functions

[johnno1962]

Values of CGFloat.native need to be cast to Double before being passed to math functions. This PR is in preparation for the Android port of Foundation in PR #622. There is a question about whether this conversion should be necessary however as CGFloat.swift is derived from code that builds on other 32 bit platforms as is discussed in the Foundation for Android PR. It’s certainly required on Android.

[Coeur]

By curiosity, it couldn't be CGFloat(acos(x.double)) instead of CGFloat(acos(Double(x.native)))?

[edit]
Nevermind, Parkera explained it should be like that because:

I don't like phrasing this as a type conversion hidden as a property.

[xwu]

@johnno1962, actually, this PR should not be necessary for any 32-bit platform. See:

https://github.com/apple/swift/blob/1d4a9707d6facd99b749e3bc9005e8542959ebc4/stdlib/public/Platform/tgmath.swift.gyb

This overlay renames acosf to acos, etc., so that CGFloat(acos(x.native)) is correct regardless of whether x.native is 32-bit or 64-bit. So something is wrong if the Android port won't compile, as the existing code should work.

[johnno1962]

This is what @phausler has been saying all along. So something is wrong with my/the Android swift toolchain/stdlib? Any ideas? I followed the recipe here: https://github.com/apple/swift/blob/master/docs/Android.md.. Rebuilding from scratch...

[xwu]

Hmm. I'm not in a position to try this myself, but:

• Can you follow that recipe and compile a "hello world"?
• Can you follow that recipe and compile a "hello world" with "print(acos(0.5 as Float))"?

[xwu]

If both of those work but "print(acos(CGFloat(0.5 as Float).native))" doesn't, then the code might be missing an "#if os(Android)" somewhere in CGFloat or a related file; otherwise, if one of those steps doesn't work, it's might be a similar problem with Float, the tgmath overlay, or something else in that arena...

[rintaro]

I think, this is because Glibc is not imported for Android.

@johnno1962
Could you try adding || os(Android) there?

Oh, you already tried that in #622. Never mind.

[rintaro]

(based on #632)
But, forcing to use Float for CGFloat.NativeType works as expected on my Ubuntu.
master...rintaro:cgfloat-forcefloat
If tgmath functions don't work in 32bit environment or Android, we might be missing something.

[johnno1962]

A bit of trouble with the rebuild:
/home/johnno/Swift/swift/stdlib/public/Platform/Glibc.swift:13:19: error: no such module 'SwiftGlibc'
@_exported import SwiftGlibc // Clang module
^
Had to copy glib.modulemap across from Ninja-ReleaseAssert/swift-linux-x86_64/lib/swift/linux/x86_64/glibc.modulemap, then it regenerated for Android anyway..

[johnno1962]

A little more progress. After a complete rebuild the CGFloat.swift problems have disappeared while the NSGeometry.swift casting is still required as values are being assigned from a Float to a Double on 32 bits despite tgmath adapting to the argument type.

Something like #632 is probably required.

Foundation/NSGeometry.swift:464:40: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
width = floor(aRect.size.width.native)
~~~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:468:42: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
height = floor(aRect.size.height.native)
~~~~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:472:39: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
width = ceil(aRect.size.width.native)
~~~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:476:41: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
height = ceil(aRect.size.height.native)
~~~~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:480:40: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
width = round(aRect.size.width.native)
~~~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:484:42: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
height = round(aRect.size.height.native)
~~~~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:490:36: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
minX = ceil(aRect.origin.x.native)
~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:494:36: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
minY = ceil(aRect.origin.y.native)
~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:498:44: error: binary operator '+' cannot be applied to two 'CGFloat.NativeType' (aka 'Float') operands
maxX = floor(aRect.origin.x.native + aRect.size.width.native)
~~~~~~~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~~~~~~~~~~
Foundation/NSGeometry.swift:498:44: note: expected an argument list of type '(Double, Double)'
maxX = floor(aRect.origin.x.native + aRect.size.width.native)
^
Foundation/NSGeometry.swift:502:44: error: binary operator '+' cannot be applied to two 'CGFloat.NativeType' (aka 'Float') operands
maxY = floor(aRect.origin.y.native + aRect.size.height.native)
~~~~~~~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~~~~~~~~~~~
Foundation/NSGeometry.swift:502:44: note: expected an argument list of type '(Double, Double)'
maxY = floor(aRect.origin.y.native + aRect.size.height.native)
^
Foundation/NSGeometry.swift:508:37: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
minX = floor(aRect.origin.x.native)
~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:512:37: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'
minY = floor(aRect.origin.y.native)
~~~~~~~~~~~~~~~^~~~~~
Double( )
Foundation/NSGeometry.swift:516:43: error: binary operator '+' cannot be applied to two 'CGFloat.NativeType' (aka 'Float') operands
maxX = ceil(aRect.origin.x.native + aRect.size.width.native)
~~~~~~~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~~~~~~~~~~
Foundation/NSGeometry.swift:516:43: note: expected an argument list of type '(Double, Double)'
maxX = ceil(aRect.origin.x.native + aRect.size.width.native)
^
Foundation/NSGeometry.swift:520:43: error: binary operator '+' cannot be applied to two 'CGFloat.NativeType' (aka 'Float') operands
maxY = ceil(aRect.origin.y.native + aRect.size.height.native)
~~~~~~~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~~~~~~~~~~~
Foundation/NSGeometry.swift:520:43: note: expected an argument list of type '(Double, Double)'
maxY = ceil(aRect.origin.y.native + aRect.size.height.native)
^
Foundation/NSGeometry.swift:526:37: error: cannot convert value of type 'CGFloat.NativeType' (aka 'Float') to expected argument type 'Double'

[johnno1962]

Closed in favour of #632 which builds for me on Android & #631. CGFloat.swift changes are not required after complete rebuild of toolchain from master.