Uh oh!
There was an error while loading. Please reload this page.
Make CTFE able to check for UB... - #78407
Merged
Merged
Conversation
oli-obk
commented
Oct 26, 2020
ContributorAuthor
@bors try @rust-timer queue |
rust-timer
commented
Oct 26, 2020
Collaborator
Awaiting bors try build completion |
bors
commented
Oct 26, 2020
Collaborator
⌛ Trying commit 60e41550266833277bb4f3fb94ddf2b2f5fa2fd8 with merge 3dbdd3b981f75f965ac04452739653a3d47ff0ed... |
bors
commented
Oct 26, 2020
Collaborator
☀️ Try build successful - checks-actions |
rust-timer
commented
Oct 26, 2020
Collaborator
Queued 3dbdd3b981f75f965ac04452739653a3d47ff0ed with parent 35debd4, future comparison URL. |
rust-timer
commented
Oct 26, 2020
Collaborator
Finished benchmarking try commit (3dbdd3b981f75f965ac04452739653a3d47ff0ed): comparison url. Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. Please note that if the perf results are neutral, you should likely undo the rollup=never given below by specifying Importantly, though, if the results of this run are non-neutral do not roll this PR up -- it will mask other regressions or improvements in the roll up. @bors rollup=never |
oli-obk
commented
Oct 27, 2020
ContributorAuthor
Looking good. It affected the stress tests as expected (I even got the ballpark guess of a 100% regression about right), and only caused regressions in very small perf tests in incremental due to the additional query (all the other regressions are in the millisecond region). We even got perf improvements in the cases where we don't need |
lcnr
reviewed
Oct 27, 2020
Uh oh!
There was an error while loading. Please reload this page.
oli-obkforce-pushed
the
ub_checkable_ctfe
branch
from
October 28, 2020 14:24
132b99c to
13e63f2Comparecamelid
commented
Oct 30, 2020
Member
(Added S-waiting-on-author so this isn't lost.) |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Dylan-DPC-zz
commented
Nov 28, 2020
@oli-obk any updates? |
oli-obk
commented
Nov 28, 2020
ContributorAuthor
Nope, I'll try to find some time for it next week |
oli-obkforce-pushed
the
ub_checkable_ctfe
branch
from
December 7, 2020 14:28
13e63f2 to
874dfb8Compareoli-obkforce-pushed
the
ub_checkable_ctfe
branch
from
December 7, 2020 15:26
874dfb8 to
9d03206Compareoli-obk
commented
Jan 11, 2021
ContributorAuthor
Well... let's leave a note here:
These perf regressions permit us a path forward with |
rust-log-analyzer
commented
Jan 11, 2021
Collaborator
The job Click to see the possible cause of the failure (guessed by this bot) |
bors
commented
Jan 11, 2021
Collaborator
💔 Test failed - checks-actions |
oli-obk
commented
Jan 12, 2021
ContributorAuthor
@bors r=RalfJung,pnkfelix |
bors
commented
Jan 12, 2021
Collaborator
📌 Commit 53e3a23 has been approved by |
bors
commented
Jan 12, 2021
Collaborator
bors
commented
Jan 12, 2021
Collaborator
☀️ Test successful - checks-actions |
This was referenced Jan 12, 2021
| | "pointer-to-integer cast" needs an rfc before being allowed inside constants | ||
| | | ||
| = note: The rules on what exactly is undefined behavior aren't clear, so this check might be overzealous. Please open an issue on the rustc repository if you believe it should not be considered undefined behavior. | ||
| note: the lint level is defined here |
Member
There was a problem hiding this comment.
@oli-obk any idea where all these lint changes are coming from? These tests were carefully crafted to exhibit certain code paths in the validator, and it seems like they are not hitting those code paths any more?
ContributorAuthor
There was a problem hiding this comment.
well... they are not hitting those code paths anymore, because we optimize the code less, so the const evaluator directly hits the UB itself, thus we don't get to the validator anymore because we error out before that.
Member
There was a problem hiding this comment.
That is very strange, why are optimizations needed to hit these code paths?
ContributorAuthor
There was a problem hiding this comment.
Probably because of the reborrows? There are not that many differences between optimized mir and unoptimized mir: https://play.rust-lang.org/?version=nightly&mode=debug&edition=2018&gist=19dfcdc8284fd5eb7c9110a6ffcade88
ContributorAuthor
There was a problem hiding this comment.
Member
There was a problem hiding this comment.
Hm, yeah, reborrows would check dereferencability...
I am concerned about those code paths in the validator bitrotting, but I don't have an idea for what to do about it. We have to somehow make sure no reborrow is inserted after the transmute...
Uh oh!
There was an error while loading. Please reload this page.
| } | ||
| Ok(()) | ||
| }; | ||
| match tcx.hir().body_const_context(def_id.expect_local()) { |
There was a problem hiding this comment.
The expect_local call here means that write_mir_pretty no longer works for non local def_ids. The ability to print out MIR for non local functions this way is critical for the MIRAI project.
Member
There was a problem hiding this comment.
Is the fix as simple as adding if let Some(local) = def_id.expect_local()? If so, are you interested in making a PR for that change?
There was a problem hiding this comment.
I doubt that the fix is that simple. I need a way of actually getting the MIR of non local functions. Perhaps use as_local and keep the old code for the None case?
ContributorAuthor
There was a problem hiding this comment.
I think what we can do is to check is_const_fn_raw for the const fn path, and use tcx.instance_mir(ty::InstanceDef::Item(def_id)) for the other two paths
Dylan-DPC-zz pushed a commit
to Dylan-DPC-zz/rust
that referenced
this pull request
Jun 16, 2021
Add a regression test for issue-76510 Fixed by rust-lang#78407, closesrust-lang#76510 r? `@oli-obk`
JohnTitor added a commit
to JohnTitor/rust
that referenced
this pull request
Jun 16, 2021
Add a regression test for issue-76510 Fixed by rust-lang#78407, closesrust-lang#76510 r? ``@oli-obk``
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
14 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
... by not doing any optimizations on the
const fnMIR used in CTFE. This means we duplicate allconst fn's MIR now, once for CTFE, once for runtime. This PR is for checking the perf effect, so we have some data when talking about https://github.com/rust-lang/const-eval/blob/master/rfcs/0000-const-ub.mdTo do this, we now have two queries for obtaining mir:
optimized_mirandmir_for_ctfe. It is now illegal to invokeoptimized_mirto obtain the MIR of a const/static item's initializer, an array length, an inline const expression or an enum discriminant initializer. Forconst fn, bothoptimized_mirandmir_for_ctfework, the former returning the MIR that LLVM should use if the function is called at runtime. Similarly it is illegal to invokemir_for_ctfeon regular functions.This is all checked via appropriate assertions and I don't think it is easy to get wrong, as there should be no
mir_for_ctfecalls outside the const evaluator or metadata encoding. Almost all rustc devs should keep usingoptimized_mir(orinstance_mirfor that matter).