[Swift CI] Build Failure: 0. OSS - Swift Incremental RA - Ubuntu 15.10 (master) #3818

gribozavr · March 30, 2016, 11:34pm

Looks like it was

github.com/apple/swift-corelibs-foundation

[SR-929] Implement NSString.init() with locale

apple:master ← seabaylea:nsstring-init

opened 12:58PM - 15 Mar 16 UTC

seabaylea

+41 -1

[SR-929](https://bugs.swift.org/browse/SR-929) covers the need to implement the …following NSString API: ``` swift public convenience init(format: String, locale: AnyObject?, arguments argList: CVaListPointer) ``` This PR adds an implementation that: - Allows the for `locale` parameter to be a `NSLocale`, `NSDictionary` or `nil`. The parameter type is checked, checking `NSLocale` first as its the expected type - Casts the type to `CFDictionary` for use in `CFString` Here I've used `unsafeBitCast` as casting to a CFDictionary directly doesn't work on Linux and there doesn't appear to be a "safer" was of casting that works. The underlying functions check to see if the actual type is a `CFLocale` or a `CFDictionary` and handles as appropriate. - Throws a `fatalError()` for invalid locale parameters The actual behaviour is undefined by the spec. As the init function isn't failable there needs to be some kind of error, and it looked better to fail here with a relevant message than further down the call stack. The behaviour is the same as for Objective-C on Mac, except for the error handling. Let me know if there's any suggest changes to the implementation. I've also added 4 tests for the implementation: 1. Checking the use of `nil` 2. Checking with the English locale using NSLocale, returning 1,000 (42.0) 3. Checking with the German locale using NSLocale, returning 1.000 (42,0) 4. Checking with a custom decimal separator using a NSDictionary, returning 1000 (42&0)

Dmitri

···

--
main(i,j){for(i=2;;i++){for(j=2;j<i;j++){if(!(i%j)){j=0;break;}}if
(j){printf("%d\n",i);}}} /*Dmitri Gribenko <gribozavr@gmail.com>*/

Philippe_Hausler · March 30, 2016, 11:43pm

pushed a fix for the test; it is likely that the ICU4C implementation has a slight difference between the Apple ICU and the one for linux.

···

On Mar 30, 2016, at 4:34 PM, Dmitri Gribenko via swift-dev <swift-dev@swift.org> wrote:

Looks like it was [SR-929] Implement NSString.init() with locale by seabaylea · Pull Request #286 · apple/swift-corelibs-foundation · GitHub

Dmitri

--
main(i,j){for(i=2;;i++){for(j=2;j<i;j++){if(!(i%j)){j=0;break;}}if
(j){printf("%d\n",i);}}} /*Dmitri Gribenko <gribozavr@gmail.com <mailto:gribozavr@gmail.com>>*/
_______________________________________________
swift-dev mailing list
swift-dev@swift.org
https://lists.swift.org/mailman/listinfo/swift-dev

Chris_Bailey · March 31, 2016, 11:07am

I'll have a dig into the ICU implementations to see if I can verify that
its the source of the differences.

Philippe:
Your fix to the test changes the expected localized form of the thousand
separator to nothing (ie, "1000") in all cases. Doesn't this break the OSX
builds as the result does become localized? We could fix it for both cases
(temporarily) by just removing the thousand and matching on the decimal
separator only.

Chris

···

From: Philippe Hausler via swift-dev <swift-dev@swift.org>
To: Dmitri Gribenko <gribozavr@gmail.com>
Cc: swift-dev <swift-dev@swift.org>
Date: 31/03/2016 00:43
Subject: Re: [swift-dev] [Swift CI] Build Failure: 0. OSS - Swift
Incremental RA - Ubuntu 15.10 (master) #3818
Sent by: swift-dev-bounces@swift.org

pushed a fix for the test; it is likely that the ICU4C implementation has
a slight difference between the Apple ICU and the one for linux.

On Mar 30, 2016, at 4:34 PM, Dmitri Gribenko via swift-dev < swift-dev@swift.org> wrote:

Looks like it was

github.com/apple/swift-corelibs-foundation

[SR-929] Implement NSString.init() with locale

apple:master ← seabaylea:nsstring-init

opened 12:58PM - 15 Mar 16 UTC

seabaylea

+41 -1

[SR-929](https://bugs.swift.org/browse/SR-929) covers the need to implement the …following NSString API: ``` swift public convenience init(format: String, locale: AnyObject?, arguments argList: CVaListPointer) ``` This PR adds an implementation that: - Allows the for `locale` parameter to be a `NSLocale`, `NSDictionary` or `nil`. The parameter type is checked, checking `NSLocale` first as its the expected type - Casts the type to `CFDictionary` for use in `CFString` Here I've used `unsafeBitCast` as casting to a CFDictionary directly doesn't work on Linux and there doesn't appear to be a "safer" was of casting that works. The underlying functions check to see if the actual type is a `CFLocale` or a `CFDictionary` and handles as appropriate. - Throws a `fatalError()` for invalid locale parameters The actual behaviour is undefined by the spec. As the init function isn't failable there needs to be some kind of error, and it looked better to fail here with a relevant message than further down the call stack. The behaviour is the same as for Objective-C on Mac, except for the error handling. Let me know if there's any suggest changes to the implementation. I've also added 4 tests for the implementation: 1. Checking the use of `nil` 2. Checking with the English locale using NSLocale, returning 1,000 (42.0) 3. Checking with the German locale using NSLocale, returning 1.000 (42,0) 4. Checking with a custom decimal separator using a NSDictionary, returning 1000 (42&0)

Dmitri

--
main(i,j){for(i=2;;i++){for(j=2;j<i;j++){if(!(i%j)){j=0;break;}}if
(j){printf("%d\n",i);}}} /*Dmitri Gribenko <gribozavr@gmail.com>*/
_______________________________________________
swift-dev mailing list
swift-dev@swift.org
https://lists.swift.org/mailman/listinfo/swift-dev
_______________________________________________
swift-dev mailing list
swift-dev@swift.org
https://lists.swift.org/mailman/listinfo/swift-dev