stdlib: Add withUnsafeBytes(of:) and withUnsafePointer(to:) for immutable arguments. #15608

jckarter · 2018-03-29T21:14:41Z

Since the functions produce pointers with tightly-scoped lifetimes there's no formal reason these have to only work on inout things. Now that arguments can be +0, we can even do this without copying values that we already have at +0.

Proposal: swiftlang/swift-evolution#822

jckarter · 2018-03-29T21:14:48Z

@swift-ci Please test

swift-ci · 2018-03-29T21:27:23Z

Build failed
Swift Test OS X Platform
Git Sha - 65457d3877ae24da0683ceba0e523fa658312b8f

natecook1000

This is great, Joe!

natecook1000 · 2018-03-29T22:30:58Z

stdlib/public/core/LifetimeManager.swift

Lost the end of the sentence here…

natecook1000 · 2018-03-29T22:33:44Z

stdlib/public/core/UnsafeRawBufferPointer.swift.gyb

This parameter is named value, and we've got lots of args above—what do you think of standardizing on value throughout? It's less jargony.

natecook1000 · 2018-03-29T22:38:35Z

stdlib/public/core/LifetimeManager.swift

Can you add the parameters to the function name where this is mentioned? withUnsafeMutablePointer(to:_:)

swift-ci · 2018-03-29T22:48:03Z

Build failed
Swift Test Linux Platform
Git Sha - 65457d3877ae24da0683ceba0e523fa658312b8f

atrick · 2018-03-30T00:20:45Z

Very nice.

jckarter · 2018-03-30T20:42:02Z

@swift-ci Please test

jckarter · 2018-03-30T20:42:16Z

@natecook1000 Thanks for the feedback! How do the doc comments look now?

swift-ci · 2018-03-30T20:42:30Z

Build failed
Swift Test OS X Platform
Git Sha - 65457d3877ae24da0683ceba0e523fa658312b8f

swift-ci · 2018-03-30T20:43:42Z

Build failed
Swift Test Linux Platform
Git Sha - 65457d3877ae24da0683ceba0e523fa658312b8f

The resilience model forces inlinable code to use resiliently conservative access paths for properties in the same module, but some stdlib methods by design escapes storage of a global variable. The additional `Builtin.addressof` validation introduced in swiftlang#15608 exposed the latent issue here.

natecook1000 · 2018-03-30T21:08:20Z

This looks great, @jckarter — thanks! Good note on exclusivity concerns, too. 👍

jckarter · 2018-03-30T21:50:36Z

@xedin Any idea why this change would have regressed diagnostics in this test in test/Constraints/diagnostics.swift?

func read<T : BinaryInteger>() -> T? {
  var buffer : T 
  let n = withUnsafePointer(to: &buffer) { (p) in
    read2(UnsafePointer(p), maxLength: MemoryLayout<T>.size) // expected-error {{cannot convert value of type 'UnsafePointer<_>' to expected argument type 'UnsafeMutableRawPointer'}}
  }
}

I would've expected no impact, since the & should still make which withUnsafePointer overload gets chosen unambiguous, but it changes the diagnostics we report, possibly due to the overload set in the solution:

******************** TEST 'Swift(macosx-x86_64) :: Constraints/diagnostics.swift' FAILED ********************
16:43:47 Script:
16:43:47 --
16:43:47 /Users/buildnode/jenkins/workspace/swift-PR-osx/branch-master/buildbot_incremental/swift-macosx-x86_64/bin/swiftc -frontend -target x86_64-apple-macosx10.9  -module-cache-path '/var/folders/_8/79jmzf2142z2xydc_01btlx00000gn/T/swift-testsuite-clang-module-cachegSRGYq' -sdk /Applications/Xcode-beta.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.13.sdk -swift-version 3  -typecheck -verify -disable-objc-attr-requires-foundation-module /Users/buildnode/jenkins/workspace/swift-PR-osx/branch-master/swift/test/Constraints/diagnostics.swift
16:43:47 --
16:43:47 Exit Code: 1
16:43:47 
16:43:47 Command Output (stderr):
16:43:47 --
16:43:47 
/Users/buildnode/jenkins/workspace/swift-PR-osx/branch-master/swift/test/Constraints/diagnostics.swift:755:11: error: unexpected note produced: explicitly specify the generic arguments to fix this issue
16:43:47     read2(UnsafePointer(p), maxLength: MemoryLayout<T>.size) // expected-error {{cannot convert value of type 'UnsafePointer<_>' to expected argument type 'UnsafeMutableRawPointer'}}
16:43:47           ^
16:43:47 /Users/buildnode/jenkins/workspace/swift-PR-osx/branch-master/swift/test/Constraints/diagnostics.swift:755:82: error: incorrect message found
16:43:47     read2(UnsafePointer(p), maxLength: MemoryLayout<T>.size) // expected-error {{cannot convert value of type 'UnsafePointer<_>' to expected argument type 'UnsafeMutableRawPointer'}}
16:43:47                                                                                  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
16:43:47                                                                                  generic parameter 'Pointee' could not be inferred
16:43:47 
16:43:47 --
16:43:47 
16:43:47 ********************

jrose-apple · 2018-03-30T22:17:05Z

stdlib/public/core/StringObject.swift

I'm pretty sure that's no good either, because we want this to fold down to a single address reference. Did we ever implement the thing where _emptyStringStorage was itself inlinable, meaning that you could rely on it being stored?

jrose-apple · 2018-03-30T22:17:42Z

stdlib/public/core/VarArgs.swift

Same here. We're supposed to fold all the va_list-building down to primitive ops.

jckarter · 2018-04-18T02:49:25Z

@swift-ci Please test

jckarter · 2018-04-18T02:49:36Z

@swift-ci Please test source compatibility

swift-ci · 2018-04-18T02:49:55Z

Build failed
Swift Test OS X Platform
Git Sha - 670903b39ede9b42ab713e4524fbf783ecb148d0

swift-ci · 2018-04-18T02:51:09Z

Build failed
Swift Test Linux Platform
Git Sha - 670903b39ede9b42ab713e4524fbf783ecb148d0

…able arguments. Since the functions produce pointers with tightly-scoped lifetimes there's no formal reason these have to only work on `inout` things. Now that arguments can be +0, we can even do this without copying values that we already have at +0.

jckarter · 2018-04-18T16:14:39Z

@swift-ci Please test

swift-ci · 2018-04-18T16:16:27Z

Build failed
Swift Test OS X Platform
Git Sha - a3536a6a541ba93f2b3dbeefe3d41317d6bf4877

swift-ci · 2018-04-18T16:16:36Z

Build failed
Swift Test Linux Platform
Git Sha - a3536a6a541ba93f2b3dbeefe3d41317d6bf4877

natecook1000 reviewed Mar 29, 2018

View reviewed changes

jckarter force-pushed the withUnsafePointer-for-lets branch from 65457d3 to 670903b Compare March 30, 2018 20:41

jckarter mentioned this pull request Mar 30, 2018

stdlib: Fix some @_inlineable/Builtin.addressof bad interactions. #15632

Merged

jckarter mentioned this pull request Mar 30, 2018

[4.2] stdlib: Fix some @_inlineable/Builtin.addressof bad interactions. #15634

Merged

jrose-apple reviewed Mar 30, 2018

View reviewed changes

jckarter force-pushed the withUnsafePointer-for-lets branch from 670903b to a3536a6 Compare April 18, 2018 02:49

jckarter force-pushed the withUnsafePointer-for-lets branch from a3536a6 to 262137d Compare April 18, 2018 16:13

jckarter force-pushed the withUnsafePointer-for-lets branch from 262137d to 62771a0 Compare April 18, 2018 16:14

jckarter merged commit 0ee381e into swiftlang:master Apr 18, 2018

jckarter changed the title ~~[pending] stdlib: Add withUnsafeBytes(of:) and withUnsafePointer(to:) for immutable arguments.~~ stdlib: Add withUnsafeBytes(of:) and withUnsafePointer(to:) for immutable arguments. Apr 18, 2018

stdlib: Add withUnsafeBytes(of:) and withUnsafePointer(to:) for immutable arguments. #15608

stdlib: Add withUnsafeBytes(of:) and withUnsafePointer(to:) for immutable arguments. #15608

Uh oh!

Conversation

jckarter commented Mar 29, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jckarter commented Mar 29, 2018

Uh oh!

swift-ci commented Mar 29, 2018

Uh oh!

natecook1000 left a comment

Choose a reason for hiding this comment

Uh oh!

natecook1000 Mar 29, 2018

Choose a reason for hiding this comment

Uh oh!

natecook1000 Mar 29, 2018

Choose a reason for hiding this comment

Uh oh!

natecook1000 Mar 29, 2018

Choose a reason for hiding this comment

Uh oh!

swift-ci commented Mar 29, 2018

Uh oh!

atrick commented Mar 30, 2018

Uh oh!

jckarter commented Mar 30, 2018

Uh oh!

jckarter commented Mar 30, 2018

Uh oh!

swift-ci commented Mar 30, 2018

Uh oh!

swift-ci commented Mar 30, 2018

Uh oh!

natecook1000 commented Mar 30, 2018

Uh oh!

jckarter commented Mar 30, 2018

Uh oh!

jrose-apple Mar 30, 2018

Choose a reason for hiding this comment

Uh oh!

jrose-apple Mar 30, 2018

Choose a reason for hiding this comment

Uh oh!

jckarter commented Apr 18, 2018

Uh oh!

jckarter commented Apr 18, 2018

Uh oh!

swift-ci commented Apr 18, 2018

Uh oh!

swift-ci commented Apr 18, 2018

Uh oh!

jckarter commented Apr 18, 2018

Uh oh!

swift-ci commented Apr 18, 2018

Uh oh!

swift-ci commented Apr 18, 2018

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

jckarter commented Mar 29, 2018 •

edited

Loading