desugaring-based box placement-in (take-4 branch) by pnkfelix · Pull Request #26180 · rust-lang/rust

pnkfelix · 2015-06-10T15:33:51Z

Macro-desugaring of box and placement-box into "simpler" expressions.

Includes Placer API that one can override to integrate support for box into one's own type.

Does not turn on the in <place-expr> { <block> ... } syntax proposed in rust-lang/rfcs#809 , so there is still more to be done in that area. But I wanted to get this into the review pipeline now, since it is more invasive than we originally anticipated it being.

rust-highfive · 2015-06-10T15:34:04Z

r? @sfackler

(rust_highfive has picked a reviewer for you, use r? to override)

pnkfelix · 2015-06-10T15:36:50Z

r? @nikomatsakis

Feel free to reassign.

Also, you've already looked at much of this before, so feel free to ask me to point out the bits that are most different. Actually, let me try to do some of that off the bat; the newest parts of this are:

pnkfelix@fe09edb : push_unsafe! and pop_unsafe! for "unsafety hygiene" (the macros themselves are not used by the box protocol -- the macros just illustrate/test the functionality that is used by the box protocol)
pnkfelix@a97598e Change signature for move_val_init intrinsic to take *mut T
pnkfelix@379a252 Revise lang_item demo to something unrelated to Box impl (which should probably be looked at by @steveklabnik independently)

bors · 2015-06-18T21:34:47Z

☔ The latest upstream changes (presumably #26192) made this pull request unmergeable. Please resolve the merge conflicts.

update test/compile-fail/feature-gate-box-expr.rs to reflect new feature gates. Part of what lands with Issue 22181.

See discussion on Issue 22231.

Namely: * Update run-pass/new-box-syntax * Fix doc-embedded test for `alloc::boxed` to reflect new syntax. * Fix test/debuginfo/box.rs to reflect new syntax. Part of what lands with Issue 22181.

…t suite. Precursor for overloaded-`box` and placement-`in`; see Issue 22181.

Precursor for overloaded-`box` and placement-`in`; see Issue 22181.

… depth. This is a total hack; it not only needs a feature-gate, but probably should be feature-gated forever (if possible). ignore-pretty in test/run-pass/pushpop-unsafe-okay.rs

…dest`. rebase update to typeck/check/mod.rs

…init`.

…pansions.

in effort to understand treatment of `allow_internal_unstable`.

…nce.

…ions.

…rotocol.

…r box protocol.

…box protocol.

… box protocol. Actually for some reason I opted for type-annotations for most of this. (And that's really what I should have been striving for in the other commits, but its just easier to write `Box::new` most of the time.)

…box protocol.

… protocol.

…boxed`.

Precursor for landing overloaded-box, since that will decouple the box syntax from the exchange heap (and should eliminate the use of the `malloc` and `free` lang items). (This is a simplified approach to PR rust-lang#22499; note that I have once again changes which lang item to use for the illustration.)

bors · 2015-07-18T03:43:12Z

☔ The latest upstream changes (presumably #26955) made this pull request unmergeable. Please resolve the merge conflicts.

nikomatsakis · 2015-07-20T19:29:38Z

@pnkfelix let's talk about this branch; I liked the idea of isolating out the placement new work and landing that first to avoid box regressions

pnkfelix · 2015-07-21T16:21:20Z

@nikomatsakis yeah I have just finished factoring the in code into such a branch. I'll put up an alternative PR soon and close this one.

nikomatsakis · 2015-07-21T16:52:30Z

Nice.

Niko

-------- Original message --------

From: Felix S Klock II notifications@github.com

Date:07/21/2015 12:21 (GMT-05:00)

To: rust-lang/rust rust@noreply.github.com

Cc: Niko Matsakis niko@alum.mit.edu

Subject: Re: [rust] desugaring-based box placement-in (take-4 branch) (#26180)

@nikomatsakis yeah I have just finished factoring the in code into such a branch. I'll put up an alternative PR soon and close this one.

—
Reply to this email directly or view it on GitHub.

pnkfelix · 2015-07-22T13:40:26Z

closing in favor of PR #27215.

…akis Macro desugaring of `in PLACE { BLOCK }` into "simpler" expressions following the in-development "Placer" protocol. Includes Placer API that one can override to integrate support for `in` into one's own type. (See [RFC 809].) [RFC 809]: https://github.com/rust-lang/rfcs/blob/master/text/0809-box-and-in-for-stdlib.md Part of #22181 Replaced PR #26180. Turns on the `in PLACE { BLOCK }` syntax, while leaving in support for the old `box (PLACE) EXPR` syntax (since we need to support that at least until we have a snapshot with support for `in PLACE { BLOCK }`. (Note that we are not 100% committed to the `in PLACE { BLOCK }` syntax. In particular I still want to play around with some other alternatives. Still, I want to get the fundamental framework for the protocol landed so we can play with implementing it for non `Box` types.) ---- Also, this PR leaves out support for desugaring-based `box EXPR`. We will hopefully land that in the future, but for the short term there are type-inference issues injected by that change that we want to resolve separately.

rust-highfive assigned sfackler Jun 10, 2015

rust-highfive assigned nikomatsakis and unassigned sfackler Jun 10, 2015

pnkfelix added 24 commits July 16, 2015 11:54

Added Box<_> type annotations where necessary for overloaded-box.
6a82cca

prototype Placer protocol for unstable overloaded-box and placement-in.
1d86250

Add feature-gates for desugaring-based box and placement-in.
f773341
update test/compile-fail/feature-gate-box-expr.rs to reflect new feature gates. Part of what lands with Issue 22181.

Accommodate error message explosion from box-desugaring in some tests.
ca73eb8
See discussion on Issue 22231.

Update tests for desugaring box and placement-in.
a7ef6f4
Namely: * Update run-pass/new-box-syntax * Fix doc-embedded test for `alloc::boxed` to reflect new syntax. * Fix test/debuginfo/box.rs to reflect new syntax. Part of what lands with Issue 22181.

Issue 22450: Address desugaring-box problems in [pretty] run-pass tes…
a807ef4
…t suite. Precursor for overloaded-`box` and placement-`in`; see Issue 22181.

Workaround issue 22462 by moving static value into its own module.
fa0ec8c
Precursor for overloaded-`box` and placement-`in`; see Issue 22181.

update alloc::boxed post rebase.
6988850

update liballoc/lib.rs post rebase.
3ec60e8

Hack for "unsafety hygiene" -- push_unsafe! and pop_unsafe! track…
6243ce3
… depth. This is a total hack; it not only needs a feature-gate, but probably should be feature-gated forever (if possible). ignore-pretty in test/run-pass/pushpop-unsafe-okay.rs

Change signature for move_val_init intrinsic to take *mut T for `…
0530a53
…dest`. rebase update to typeck/check/mod.rs

Avoid feature-warnings on stage0.
c364f04

Revise placement-in expansion to use push/pop_unsafe and `move_val_…
70343c8
…init`.

call push_compiler_expansion during the box and placement-in ex…
69056cb
…pansions.

turn on use of sugaring based box EXPR in parser.
3ab0ebe

Instrument rustc::middle::stability::maybe_do_stability_checking
bfe8884
in effort to understand treatment of `allow_internal_unstable`.

Add Box::new in spots where new box protocol needs help with infere…
74676b2
…nce.

Allow unstable code to be injected by box and placement-in expans…
a8ef3c4
…ions.

librustc: Replace box with Box::new to assist inference for box p…
5d805b0
…rotocol.

librustc_driver: Replace box with Box::new to assist inference fo…
4d1635d
…r box protocol.

librustc_lint: Replace box with Box::new to assist inference for …
7d77257
…box protocol.

librustc_llvm: Replace box with Box::new to assist inference for …
52229f7
…box protocol.

librustdoc: Replace box with Box::new to assist inference for box…
91cf845
… protocol.

pnkfelix added 12 commits July 16, 2015 14:44

switch to box expr form in liblog.
19020ef

Add Box<_> annotations to run-pass tests for box protocol.
fa89d84

Switched to Box::new in run-fail test.
62650f9

Accommodate box protocol in compile-fail tests.
474c00c

add some (potentially unnecessary) #[inline] annotations in alloc::…
30db8cb
…boxed`.

accommodate box protocol in run-pass-valgrind.
36ebfeb

Remove dependences on box-protocol from run-pass-fulldeps.
6caa526

Fix box-protocol for run-make tests.
4d80e03

Placate tidy, mostly in test suite code.
1b8813d

Placate box protocol by switching to Box::new in various test code.
ef6d5e3

revert back to the old syntax while landing the protocol alone.
a61490c

pnkfelix force-pushed the fsk-box-placer-take-4 branch from dca3dcf to a61490cCompare July 16, 2015 12:45

pnkfelix mentioned this pull request Jul 22, 2015
desugaring-based placement-in (just in, no box); take-5 branch #27215
Merged

pnkfelix closed this Jul 22, 2015

Navigation Menu

Uh oh!

Uh oh!

desugaring-based box placement-in (take-4 branch) - #26180

desugaring-based box placement-in (take-4 branch)#26180
pnkfelix wants to merge 36 commits into
rust-lang:masterfrom
pnkfelix:fsk-box-placer-take-4

pnkfelix commented Jun 10, 2015

Uh oh!

rust-highfive commented Jun 10, 2015

Uh oh!

pnkfelix commented Jun 10, 2015

Uh oh!

bors commented Jun 18, 2015

Uh oh!

bors commented Jul 18, 2015

Uh oh!

nikomatsakis commented Jul 20, 2015

Uh oh!

pnkfelix commented Jul 21, 2015

Uh oh!

nikomatsakis commented Jul 21, 2015

Uh oh!

pnkfelix commented Jul 22, 2015

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Uh oh!

Uh oh!

Conversation

pnkfelix commented Jun 10, 2015

Uh oh!

rust-highfive commented Jun 10, 2015

Uh oh!

pnkfelix commented Jun 10, 2015

Uh oh!

bors commented Jun 18, 2015

Uh oh!

bors commented Jul 18, 2015

Uh oh!

nikomatsakis commented Jul 20, 2015

Uh oh!

pnkfelix commented Jul 21, 2015

Uh oh!

nikomatsakis commented Jul 21, 2015

Uh oh!

pnkfelix commented Jul 22, 2015

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants