Use rustc generated shims instead of interpreter hacks #156

oli-obk · 2017-03-23T14:46:28Z

fixes #119

~~Unfortunately I broke some things:~~

~~1904 success, 197 no mir, 206 crate not found, 71 failed, 46 C fn, 0 ABI, 121 unsupported, 6 intrinsic~~
~~1910 success, 197 no mir, 206 crate not found, 64 failed, 46 C fn, 0 ABI, 122 unsupported, 6 intrinsic~~
~~1918 success, 197 no mir, 206 crate not found, 56 failed, 46 C fn, 0 ABI, 122 unsupported, 6 intrinsic~~
~~1921 success, 199 no mir, 206 crate not found, 51 failed, 46 C fn, 0 ABI, 122 unsupported, 6 intrinsic~~
1925 success, 200 no mir, 206 crate not found, 46 failed, 46 C fn, 0 ABI, 122 unsupported, 6 intrinsic

  14 main function not found
  12 a raw memory access tried to access part of a pointer value as raw bytes
   9 the evaluated program panicked
   2 attempted to do math or a comparison on pointers into different allocations
   2 Overflow(Add) at /checkout/src/libcore/ops.rs:263:45: 263:57
   1 unused import: `std::option`
   1 tried to call a function with sig extern "C" fn(usize) -> Foo through a function pointer of type extern "C" fn(usize) -> u32
   1 tried to call a function with sig extern "C" fn(&isize) -> std::option::Option<&isize> through a function pointer of type extern "C" fn(&isize) -> &isize
   1 tried to access memory with alignment 1, but alignment 8 is required
   1 src/step.rs:250: static def id doesn't point to item
   1 internal compiler 
   1 Overflow(Neg) at /checkout/src/libcore/ops.rs:758:43: 758:45

oli-obk · 2017-03-23T14:49:48Z

And travis broke due to travis moving to rustup 🎉

eddyb · 2017-03-23T16:01:09Z

src/terminator/intrinsic.rs

                let env = self.tcx.empty_parameter_environment();
                let needs_drop = self.tcx.type_needs_drop_given_env(ty, &env);
                self.write_primval(dest, PrimVal::from_bool(needs_drop), dest_ty)?;
            }

            "offset" => {
-                let pointee_ty = substs.type_at(0);
+                let pointee_ty = instance.substs.type_at(0);


I feel like all of these wouldn't need to be changed if you copied substs out. It's just a reference anyway.

Yea I want to do some cleanups after everything works.

Do you still plan to do some cleanups? Should I wait until those are done before reviewing?

Not in this PR. I wanted to separate it from any cleanup noise

IMO it's a very simple change in this function that cleans up the diff.

dwrensha · 2017-03-23T17:07:59Z

For the sake of quick cross-reference, here's a link to the copied-in code from rustc_trans::monomorphize: monomorphize.rs

oli-obk · 2017-03-23T17:44:57Z

Ok... I fixed all the fallout.

We have one new Overflow(Add)

full log

@arielb1's changes fixed all 3 of our Type parameter T/#0(T/0) out of range bugs

solson · 2017-04-20T01:48:31Z

Sorry for the delay. I've added @eddyb and @oli-obk as collaborators on the repo so you two can make progress without blocking on me. Unfortunately, I don't have much free time/energy these days to spend on this. That said, I'll still be following the progress and reviewing as much as I can. I'll also be looking for ways to fit this back into my life again, but I shouldn't make any promises.

Feel free to either commit to master or continue doing PRs - I'll try to keep pace either way. Just don't consider my approval a requirement from now on.

eddyb · 2017-04-20T06:36:22Z

@oli-obk Can you update to the latest nightly or should I try?

EDIT: I restarted Travis which should check the latest nightly.

eddyb · 2017-04-20T06:44:35Z

Unable to update https://github.com/quininer/byteorder.git?branch=i128#ef51df29

Probably needs to be switched to 9bab6d7783f81da50feb234a120c918d9eabba6e now, or just the branch left in, if that's possible (it's a feature branch so they force pushed).

oli-obk · 2017-04-20T07:56:11Z

Won't get to this before tuesday

oli-obk · 2017-04-20T07:57:00Z

You need to change the got path to the official master branch. We are on an unofficial repo right now

eddyb · 2017-04-20T07:58:52Z

@oli-obk Did the i128 branch get merged or something? AFAICT miri needs the i128 support. So I updated the hash to the top of the branch (hope you don't mind me committing on your master - you should use a different branch because @solson and me get to modify your master branch :))).

eddyb · 2017-04-21T10:34:27Z

@oli-obk tests/compile-fail/never_say_never.rs looks like it now enters an infinite loop?

oli-obk added 18 commits March 21, 2017 09:21

Rustup to rustc 1.17.0-nightly (134c4a0f0 2017-03-20)

c6a18ce

Compiles again

dc1b0fb

Closures work again

26c3335

Fix drop terminator

030f00a

Roll our own MIR for dropping arrays.

7c12ebc

Dropping arrays works again

9e4e6cd

Dropping trait objects works again

c409079

Dropping slices works again

3ef0b0d

Refactor drop into its own module and fix Vec

caed365

Fix function pointer calls

0255a51

Reuse more rustc code instead of copying it into miri

70ea218

Refactor function calls

d70b79c

Fix virtual function calls

ad4f6b9

Fix virtual function calls

d71f24c

Fix manual rust-call impls

4e83659

Reintroduce fn ptr transmute check

1391c5a

Skip the transmute checks for closure glue

1c9f5ac

Merge branch 'master' of https://github.com/tsion/miri

1263d56

oli-obk added 4 commits March 23, 2017 15:52

Reduce noisyness

eecc727

Try to fix travis

197226a

Fix single field by val tuples

fb7cc3c

Add regression test for single field by val tuples

0927829

eddyb reviewed Mar 23, 2017

View reviewed changes

oli-obk added 2 commits March 23, 2017 17:36

print doesn't add a stack frame, so don't write arguments

f4ed482

Fix casting generic functions to concrete function pointers

cb867d2

Thinify the fat pointer on virtual function calls

bbeb721

Add regression test for bad substs

6706d8f

oli-obk added 2 commits March 27, 2017 10:13

Rustup to rustc 1.17.0-nightly (7846dbe0c 2017-03-26)

065e959

Cleanup the diff

fdeee8f

Update byteorder commit hash in Cargo.lock.

f95cc52

Update for changes in LocalDecl on nightly.

cb3799b

eddyb added 2 commits April 21, 2017 13:56

Handle Use of ! as Unreachable is not emitted nowadays.

738c7d2

Import EvalError in lvalue.

b9bd747

eddyb merged commit 9a32772 into rust-lang:master Apr 21, 2017

Use rustc generated shims instead of interpreter hacks #156

Use rustc generated shims instead of interpreter hacks #156

Uh oh!

Conversation

oli-obk commented Mar 23, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

oli-obk commented Mar 23, 2017

Uh oh!

eddyb Mar 23, 2017

Choose a reason for hiding this comment

Uh oh!

oli-obk Mar 23, 2017

Choose a reason for hiding this comment

Uh oh!

solson Mar 28, 2017

Choose a reason for hiding this comment

Uh oh!

oli-obk Mar 28, 2017

Choose a reason for hiding this comment

Uh oh!

eddyb Mar 28, 2017

Choose a reason for hiding this comment

Uh oh!

oli-obk Mar 29, 2017

Choose a reason for hiding this comment

Uh oh!

dwrensha commented Mar 23, 2017

Uh oh!

oli-obk commented Mar 23, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

solson commented Apr 20, 2017

Uh oh!

eddyb commented Apr 20, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

eddyb commented Apr 20, 2017

Uh oh!

oli-obk commented Apr 20, 2017

Uh oh!

oli-obk commented Apr 20, 2017

Uh oh!

eddyb commented Apr 20, 2017

Uh oh!

eddyb commented Apr 21, 2017

Uh oh!

Uh oh!

oli-obk commented Mar 23, 2017 •

edited

Loading

oli-obk commented Mar 23, 2017 •

edited

Loading

eddyb commented Apr 20, 2017 •

edited

Loading