Kill remaining uses of mem::uninitialized in libcore, liballoc #57045

RalfJung · 2018-12-21T14:53:11Z

Kill remaining uses of mem::uninitialized in libcore and liballoc, and enable a lint that will warn when uses are added again in the future.

To avoid casting raw pointers around (which is always very dangerous because it is not typechecked at all -- it doesn't even get the "same size" sanity check that transmute gets), I also added two new functions to MaybeUninit:

    /// Get a pointer to the first contained values.
    pub fn first_ptr(this: &[MaybeUninit<T>]) -> *const T {
        this as *const [MaybeUninit<T>] as *const T
    }

    /// Get a mutable pointer to the first contained values.
    pub fn first_mut_ptr(this: &mut [MaybeUninit<T>]) -> *mut T {
        this as *mut [MaybeUninit<T>] as *mut T
    }

I changed some of the existing code to use array-of-MaybeUninit instead of MaybeUninit-of-array, successfully removing raw pointer casts there as well.

rust-highfive · 2018-12-21T14:53:14Z

r? @bluss

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

nagisa · 2018-12-21T14:59:21Z

I was surprised recently to find that MaybeUninit is still unstable. Perhaps it would be a good idea to first let it go stable before migrating the whole standard library to it?

RalfJung · 2018-12-21T15:01:27Z

I think it makes a lot of sense to migrate more things to it to get some experience with the API before stabilizing.

src/liballoc/collections/btree/node.rs

cramertj · 2018-12-21T18:23:53Z

src/liballoc/collections/btree/node.rs

+            // Creating a `[MaybeUninit; N]` array by first creating a
+            // `MaybeUninit<[MaybeUninit; N]>`; the `into_inner` is safe because the inner
+            // array does not require initialization.
+            keys: MaybeUninit::uninitialized().into_inner(),


why do I have a terrible feeling I'm going to be reading MaybeUninit::uninitialized().into_inner() everywhere rather than mem::uninitialized()...

is it possible to wrap it into a MaybeUninit::uninitialized_array() -> [MaybeUninit; #] static method that has the pre-copy as a doc-comment so as to remove the 7 times that exact comment is given? or is that (blocked by) what Centril meant by #49147 ?

As I mentioned above, this can be made safe once #49147 is stable. Once that happens, you can just write [MaybeUninit::uninitialized(), N] directly.

@fbstj You could make a fn cap_uninit_array<T>() -> [MaybeUninit<T>; N] function locally here for some specific Ns (e.g. CAPACITY) using the unsafe method but you need const generics to do it for all const N: usize. So waiting for #49147 seems better.

why do I have a terrible feeling I'm going to be reading MaybeUninit::uninitialized().into_inner() everywhere rather than mem::uninitialized()...

In my defense, that's not the only change here. ;)

The main problem is that we have pretty bad support for working with arrays, and arrays of MaybeUninit are no exceptions. If we had functions generic over array lengths, this all would be much easier.

Yeah-- it's just that working with uninitialized arrays is my only current use of mem::uninitialized, and it used to use ManuallyDrop etc. but that route is now broken. Seeing this working-around of MaybeUninit's semantics makes me sad, but I understand it's a stepping-stone to the eventual goal.

RalfJung · 2018-12-22T10:03:04Z

I have added a macro to create uninitialized arrays. Unfortunately it requires repeating the array length, but there will be a compile error if you use the wrong one. This removes all the new into_inner. Do you think that is an improvement?

comex · 2018-12-22T10:38:47Z

It's too bad the names as_ptr and as_mut_ptr are already taken, since using them for this would be consistent with the correspondingly named methods on slices.

On the other hand, it's nice to be explicit.

RalfJung · 2018-12-22T10:47:54Z

It's too bad the names as_ptr and as_mut_ptr are already taken, since using them for this would be consistent with the correspondingly named methods on slices.

I actually feel like the methods on the slice should have been called differently to contain something like first or begin. When writing this patch, it happened several times that existing code that sued to call as_mut_ptr on slices now called the same-named method on MaybeUninit, leading to rather strange type error messages.

But well, too late for any of that. What I'd really like to do is make these functions methods that take self, but I don't think that is possible.

RalfJung · 2018-12-26T21:38:43Z

Rebased. However, stdsimd introduced some new uses of mem::zeroed into libcore...

RalfJung · 2018-12-30T11:55:08Z

Updated stdsimd.

RalfJung · 2019-01-01T15:31:55Z

So, how do we proceed with this?

bors · 2019-01-05T20:32:10Z

☔ The latest upstream changes (presumably #57354) made this pull request unmergeable. Please resolve the merge conflicts.

bors · 2019-01-27T20:54:50Z

☔ The latest upstream changes (presumably #57765) made this pull request unmergeable. Please resolve the merge conflicts.

bors · 2019-01-28T12:44:35Z

⌛ Testing commit 6a52ca3 with merge acec5ebdc09eed0dc05329bcfea27749b48a4ee0...

RalfJung · 2019-01-28T16:35:01Z

Fixed debug tests.

@bors r=SimonSapin

bors · 2019-01-28T16:35:01Z

📌 Commit 489a792 has been approved by SimonSapin

bors · 2019-01-28T16:55:49Z

⌛ Testing commit 489a792 with merge 8424c01...

Kill remaining uses of mem::uninitialized in libcore, liballoc Kill remaining uses of mem::uninitialized in libcore and liballoc, and enable a lint that will warn when uses are added again in the future. To avoid casting raw pointers around (which is always very dangerous because it is not typechecked at all -- it doesn't even get the "same size" sanity check that `transmute` gets), I also added two new functions to `MaybeUninit`: ```rust /// Get a pointer to the first contained values. pub fn first_ptr(this: &[MaybeUninit<T>]) -> *const T { this as *const [MaybeUninit<T>] as *const T } /// Get a mutable pointer to the first contained values. pub fn first_mut_ptr(this: &mut [MaybeUninit<T>]) -> *mut T { this as *mut [MaybeUninit<T>] as *mut T } ``` I changed some of the existing code to use array-of-`MaybeUninit` instead of `MaybeUninit`-of-array, successfully removing raw pointer casts there as well.

bors · 2019-01-28T19:57:48Z

💔 Test failed - checks-travis

rust-highfive · 2019-01-28T19:57:49Z

Your PR failed on Travis (raw log). Through arcane magic we have determined that the following fragments from the build log may contain information about the problem.

Click to expand the log.

I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact @TimNN. (Feature Requests)

Centril · 2019-01-28T20:02:06Z

@bors retry

Seems spurious.

…apin Kill remaining uses of mem::uninitialized in libcore, liballoc Kill remaining uses of mem::uninitialized in libcore and liballoc, and enable a lint that will warn when uses are added again in the future. To avoid casting raw pointers around (which is always very dangerous because it is not typechecked at all -- it doesn't even get the "same size" sanity check that `transmute` gets), I also added two new functions to `MaybeUninit`: ```rust /// Get a pointer to the first contained values. pub fn first_ptr(this: &[MaybeUninit<T>]) -> *const T { this as *const [MaybeUninit<T>] as *const T } /// Get a mutable pointer to the first contained values. pub fn first_mut_ptr(this: &mut [MaybeUninit<T>]) -> *mut T { this as *mut [MaybeUninit<T>] as *mut T } ``` I changed some of the existing code to use array-of-`MaybeUninit` instead of `MaybeUninit`-of-array, successfully removing raw pointer casts there as well.

@ghost

Rollup of 7 pull requests Successful merges: - #57045 (Kill remaining uses of mem::uninitialized in libcore, liballoc) - #57674 (Avoid erase_regions_ty queries if there are no regions to erase) - #57833 (Print a slightly clearer message when failing to launch a thread) - #57859 (Fix invalid background color) - #57904 (add typo suggestion to unknown attribute error) - #57915 (Pretty print `$crate` as `crate` or `crate_name` in more cases) - #57950 (Extend E0106, E0261) Failed merges: r? @ghost

@tromey

…Mark-Simulacrum Test gdb pretty printing more and fix overzealous type substitution Adresses a problem concerning printing BTreeMap / BTreeSet data in gdb: when the key or value type name contains substring "LeafNode", and the map has multiple nodes (e.g. more than 11 elements), printing causes an exception. E.g. ``` rustc -g - <<EOF use std::collections::BTreeMap; struct MyLeafNode(i8); fn main() { let m: BTreeMap<i8, MyLeafNode> = (0..12).map(|i| (i, MyLeafNode(i))).collect(); assert!(!m.is_empty()); } EOF ``` ``` $ rust-gdb rust_out (gdb) b 7 (gdb) r (gdb) p m $1 = BTreeMap<i8, rust_out::MyLeafNode>(len: 12)Python Exception <class 'gdb.error'> No type named alloc::collections::btree::node::InternalNode<i8, rust_out::MyInternalNode>.: use std::collections::BTreeMap; ``` The code was written in rust-lang#56144 by @tromey (and later touched upon by @RalfJung in rust-lang#57045, but I think that had nothing to do with the issues in this PR).

rust-highfive assigned bluss Dec 21, 2018

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Dec 21, 2018

RalfJung mentioned this pull request Dec 21, 2018

Tracking issue for RFC 1892, "Deprecate uninitialized in favor of a new MaybeUninit type" #53491

Closed

3 tasks

Centril reviewed Dec 21, 2018

View reviewed changes

src/liballoc/collections/btree/node.rs Outdated Show resolved Hide resolved

cramertj reviewed Dec 21, 2018

View reviewed changes

This comment has been minimized.

Sign in to view

RalfJung force-pushed the kill-more-uninit branch from 36dcd75 to 4a6497b Compare December 23, 2018 13:29

This comment has been minimized.

Sign in to view

RalfJung force-pushed the kill-more-uninit branch from 4a6497b to b9595b3 Compare December 26, 2018 21:37

RalfJung mentioned this pull request Dec 26, 2018

avoid deprecated mem::zeroed rust-lang/stdarch#628

Merged

This comment has been minimized.

Sign in to view

RalfJung force-pushed the kill-more-uninit branch from b9595b3 to 8e312e7 Compare December 30, 2018 11:54

RalfJung force-pushed the kill-more-uninit branch from 8e312e7 to f4dd732 Compare January 9, 2019 15:43

RalfJung added 3 commits January 28, 2019 10:39

libcore: remove unneeded allow(deprecated)

c775c2f

liballoc: remove unneeded allow(deprecated)

c11e514

libcore: avoid mem::uninitialized and raw ptr casts

a88414e

RalfJung force-pushed the kill-more-uninit branch from f4dd732 to 678bd35 Compare January 28, 2019 09:47

RalfJung added 2 commits January 28, 2019 10:48

avoid mem::uninitialized in BTreeMap

ffd73df

avoid some raw ptr casts in BTreeMap

630aaa4

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jan 28, 2019

This comment has been minimized.

Sign in to view

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Jan 28, 2019

fix gdb debug printing

489a792

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jan 28, 2019

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Jan 28, 2019

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jan 28, 2019

Centril mentioned this pull request Jan 28, 2019

Rollup of 7 pull requests #57957

Merged

bors merged commit 489a792 into rust-lang:master Jan 29, 2019

RalfJung deleted the kill-more-uninit branch February 17, 2019 11:06

ssomers mentioned this pull request Jan 10, 2020

Test gdb pretty printing more and fix overzealous type substitution #68098

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Kill remaining uses of mem::uninitialized in libcore, liballoc #57045

Kill remaining uses of mem::uninitialized in libcore, liballoc #57045

RalfJung commented Dec 21, 2018

rust-highfive commented Dec 21, 2018

nagisa commented Dec 21, 2018

RalfJung commented Dec 21, 2018

cramertj Dec 21, 2018

fbstj Dec 21, 2018

Centril Dec 21, 2018

Centril Dec 21, 2018

RalfJung Dec 22, 2018

cramertj Dec 26, 2018

RalfJung commented Dec 22, 2018

comex commented Dec 22, 2018

RalfJung commented Dec 22, 2018

This comment has been minimized.

This comment has been minimized.

RalfJung commented Dec 26, 2018

This comment has been minimized.

RalfJung commented Dec 30, 2018

RalfJung commented Jan 1, 2019

bors commented Jan 5, 2019

bors commented Jan 27, 2019

bors commented Jan 28, 2019

This comment has been minimized.

This comment has been minimized.

RalfJung commented Jan 28, 2019

bors commented Jan 28, 2019

bors commented Jan 28, 2019

bors commented Jan 28, 2019

rust-highfive commented Jan 28, 2019

Centril commented Jan 28, 2019

Kill remaining uses of mem::uninitialized in libcore, liballoc #57045

Kill remaining uses of mem::uninitialized in libcore, liballoc #57045

Conversation

RalfJung commented Dec 21, 2018

rust-highfive commented Dec 21, 2018

nagisa commented Dec 21, 2018

RalfJung commented Dec 21, 2018

cramertj Dec 21, 2018

Choose a reason for hiding this comment

fbstj Dec 21, 2018

Choose a reason for hiding this comment

Centril Dec 21, 2018

Choose a reason for hiding this comment

Centril Dec 21, 2018

Choose a reason for hiding this comment

RalfJung Dec 22, 2018

Choose a reason for hiding this comment

cramertj Dec 26, 2018

Choose a reason for hiding this comment

RalfJung commented Dec 22, 2018

comex commented Dec 22, 2018

RalfJung commented Dec 22, 2018

This comment has been minimized.

This comment has been minimized.

RalfJung commented Dec 26, 2018

This comment has been minimized.

RalfJung commented Dec 30, 2018

RalfJung commented Jan 1, 2019

bors commented Jan 5, 2019

bors commented Jan 27, 2019

bors commented Jan 28, 2019

This comment has been minimized.

This comment has been minimized.

RalfJung commented Jan 28, 2019

bors commented Jan 28, 2019

bors commented Jan 28, 2019

bors commented Jan 28, 2019

rust-highfive commented Jan 28, 2019

Centril commented Jan 28, 2019