platform/upstream/llvm.git
3 years ago[compiler-rt] Produce the right arch suffix for arm libraries
Martin Storsjö [Thu, 11 Mar 2021 21:44:16 +0000 (23:44 +0200)]
[compiler-rt] Produce the right arch suffix for arm libraries

If producing libraries with an arch suffix (i.e. if
LLVM_ENABLE_PER_TARGET_RUNTIME_DIR isn't set), we append the
architecture name. However, for arm, clang doesn't look for libraries
with the full architecture name, but only looks for "arm" and "armhf".

Try to deduce what the full target triple might have been, and use
that for deciding between "arm" and "armhf".

This tries to reapply this bit from D98173, that had to be reverted
in 7b153b43d3a14d76975039408c4b922beb576735 due to affecting how
the builtins themselves are compiled, not only affecting the output
file name.

Differential Revision: https://reviews.llvm.org/D98452

3 years ago[NFC] One more use case for evaluatePredicate
Max Kazantsev [Thu, 18 Mar 2021 11:50:55 +0000 (18:50 +0700)]
[NFC] One more use case for evaluatePredicate

3 years ago[NFC] Use evaluatePredicate in eliminateComparison
Max Kazantsev [Thu, 18 Mar 2021 11:48:10 +0000 (18:48 +0700)]
[NFC] Use evaluatePredicate in eliminateComparison

Just makes code simpler.

3 years ago[SCEV][NFC] API for predicate evaluation
Max Kazantsev [Thu, 18 Mar 2021 11:28:14 +0000 (18:28 +0700)]
[SCEV][NFC] API for predicate evaluation

Provides API that allows to check predicate for being true or
false with one call. Current implementation is naive and just
calls isKnownPredicate twice, but further we can rework this
logic trying to use one check to prove both facts.

3 years ago[test] Fix incorrect use of string variable use
Thomas Preud'homme [Thu, 18 Mar 2021 10:45:55 +0000 (10:45 +0000)]
[test] Fix incorrect use of string variable use

LLVM test CodeGen/AArch64/machine-outliner-retaddr-sign-thunk.ll uses
a string substitution block that contains a regex matching block. This
seems like as a copy/paste from other similar test where the match also
defines a variable, hence the [[]] syntax. In this case however this is
a CHECK-NOT variable so nothing should match. No variable definition is
thus expected and the square brackets can be dropped.

Reviewed By: chill

Differential Revision: https://reviews.llvm.org/D98853

3 years ago[OpenCL] Remove spurious atomic_fetch tablegen builtins
Sven van Haastregt [Thu, 18 Mar 2021 12:17:12 +0000 (12:17 +0000)]
[OpenCL] Remove spurious atomic_fetch tablegen builtins

The `int` and `long` versions of these builtins already provide the
necessary overloads for `intptr_t` and `uintptr_t` arguments, as
`ASTContext` defines `atomic_(u)intptr_t` in terms of the `int` or
`long` types.

Prior to this patch, calls to those builtins with particular argument
types resulted in call-is-ambiguous errors.

Differential Revision: https://reviews.llvm.org/D98520

3 years ago[test] Fix variable definition in acle_sve_ld1.sh
Thomas Preud'homme [Thu, 18 Mar 2021 10:36:15 +0000 (10:36 +0000)]
[test] Fix variable definition in acle_sve_ld1.sh

Clang test acle_sve_ld1.sh is missing the colon in one of the string
variable definition separating the variable name from the regex. This
leads the substitution block to be parsed as a numeric variable use.

Reviewed By: sdesmalen

Differential Revision: https://reviews.llvm.org/D98852

3 years ago[LoopVectorize] relax FMF constraint for FP induction
Sanjay Patel [Thu, 18 Mar 2021 12:09:28 +0000 (08:09 -0400)]
[LoopVectorize] relax FMF constraint for FP induction

This makes the induction part of the loop vectorizer match the reduction part.
We do not need all of the fast-math-flags. For example, there are some that
clearly are not in play like arcp or afn.

If we want to make FMF constraints consistent across the IR optimizer, we
might want to add nsz too, but that's up for debate (users can't expect
associative FP math and preservation of sign-of-zero at the same time?).

The calling code was fixed to avoid miscompiles with:
1bee549737ac

Differential Revision: https://reviews.llvm.org/D98708

3 years ago[clang][ASTImporter] Fix import of VarDecl regarding thread local storage spec
Balazs Benics [Thu, 18 Mar 2021 12:06:38 +0000 (13:06 +0100)]
[clang][ASTImporter] Fix import of VarDecl regarding thread local storage spec

After the import, we did not copy the `TSCSpec`.
This commit resolves that.

Reviewed By: balazske

Differential Revision: https://reviews.llvm.org/D98707

3 years ago[-Wcalled-once-parameter][NFC] Fix GCC compilation error
Valeriy Savchenko [Thu, 18 Mar 2021 11:22:45 +0000 (14:22 +0300)]
[-Wcalled-once-parameter][NFC] Fix GCC compilation error

3 years ago[Reland] "Do not apply calling conventions to MSVC entry points"
Elizabeth Andrews [Thu, 18 Mar 2021 09:58:35 +0000 (02:58 -0700)]
[Reland] "Do not apply calling conventions to MSVC entry points"

This patch is a second attempt at fixing a link error for MSVC
entry points when calling conventions are specified using a flag.

Calling conventions specified using flags should not be applied to MSVC
entry points. The default calling convention is set in this case. The
default calling convention for MSVC entry points main and wmain is cdecl.
For WinMain, wWinMain and DllMain, the default calling convention is
stdcall on 32 bit Windows.

Explicitly specified calling conventions are applied to MSVC entry points.

For MinGW, the default calling convention for all MSVC entry points is
cdecl.

First attempt: 4cff1b40dacf6
Revert of first attempt: bebfc3b92d5e8

Differential Revision: https://reviews.llvm.org/D97941

3 years ago[AMDGPU] Regenerate atomic_optimizations_global_pointer.ll tests
Simon Pilgrim [Thu, 18 Mar 2021 11:15:44 +0000 (11:15 +0000)]
[AMDGPU] Regenerate atomic_optimizations_global_pointer.ll tests

3 years ago[ARM] Regenerate select-imm.ll tests
Simon Pilgrim [Thu, 18 Mar 2021 11:07:16 +0000 (11:07 +0000)]
[ARM] Regenerate select-imm.ll tests

3 years ago[llvm-objcopy] remove split dwo file creation from executeObjcopyOnBinary.
Alexey Lapshin [Thu, 11 Mar 2021 22:31:06 +0000 (01:31 +0300)]
[llvm-objcopy] remove split dwo file creation from executeObjcopyOnBinary.

This patch removes creation of the resulting file from the
executeObjcopyOnBinary() function. For the most use cases, the
executeObjcopyOnBinary receives output file as a parameter
- raw_ostream &Out. The splitting .dwo file is implemented differently:
file containg .dwo tables is created inside executeObjcopyOnBinary().
When objcopy functionality would be moved into separate library,
current implementation will become inconvenient. The goal of that
refactoring is to separate concerns: It might be convenient to
to do dwo tables splitting but to create resulting file differently.

Differential Revision: https://reviews.llvm.org/D98582

3 years ago[DAG] SelectionDAG::isSplatValue - add ISD::ABS handling
Simon Pilgrim [Thu, 18 Mar 2021 10:26:46 +0000 (10:26 +0000)]
[DAG] SelectionDAG::isSplatValue - add ISD::ABS handling

Add ISD::ABS to the existing unary instructions handling for splat detection

This is similar to D83605, but doesn't appear to need to touch any of the wasm refactoring.

Differential Revision: https://reviews.llvm.org/D98778

3 years ago[RISCV] Support scalable-vector masked scatter operations
Fraser Cormack [Mon, 8 Feb 2021 15:33:23 +0000 (15:33 +0000)]
[RISCV] Support scalable-vector masked scatter operations

This patch adds support for masked scatter intrinsics on scalable vector
types. It is mostly an extension of the earlier masked gather support
introduced in D96263, since the addressing mode legalization is the
same.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D96486

3 years ago[Test][DebugInfo] Check for backend object emission support.
Nigel Perks [Mon, 14 Sep 2020 17:17:11 +0000 (18:17 +0100)]
[Test][DebugInfo] Check for backend object emission support.

The XCore backend does not support object emission. Several tests fail for this
reason when XCore is the default target. See staging buildbot builder:
clang-xcore-ubuntu-20-x64.

So check for backend object emission before running the tests requiring it.

Incorporate isConfigurationSupported functionality in isObjectEmissionSupported,
to avoid calling them both in the same tests.

Differential Revision: https://reviews.llvm.org/D98400

3 years ago[RISCV] Support scalable-vector masked gather operations
Fraser Cormack [Thu, 4 Feb 2021 09:56:01 +0000 (09:56 +0000)]
[RISCV] Support scalable-vector masked gather operations

This patch supports the masked gather intrinsics in RVV.

The RVV indexed load/store instructions only support the "unsigned unscaled"
addressing mode; indices are implicitly zero-extended or truncated to XLEN and
are treated as byte offsets. This ISA supports the intrinsics directly, but not
the majority of various forms of the MGATHER SDNode that LLVM combines to. Any
signed or scaled indexing is extended to the XLEN value type and scaled
accordingly. This is done during DAG combining as widening the index types to
XLEN may produce illegal vectors that require splitting, e.g.
nxv16i8->nxv16i64.

Support for scalable-vector CONCAT_VECTORS was added to avoid spilling via the
stack when lowering split legalized index operands.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D96263

3 years ago[-Wcalled-once-parameter] Fix false positives for cleanup attr
Valeriy Savchenko [Thu, 11 Mar 2021 11:22:47 +0000 (14:22 +0300)]
[-Wcalled-once-parameter] Fix false positives for cleanup attr

Cleanup attribute allows users to attach a destructor-like functions
to variable declarations to be called whenever they leave the scope.
The logic of such functions is not supported by the Clang's CFG and
is too hard to be reasoned about.  In order to avoid false positives
in this situation, we assume that we didn't see ALL of the executtion
paths of the function and, thus, can warn only about multiple call
violation.

rdar://74441906

Differential Revision: https://reviews.llvm.org/D98694

3 years ago[mlir][ODS] Get rid of limitations in rewriters generator
Vladislav Vinogradov [Wed, 3 Mar 2021 09:04:08 +0000 (12:04 +0300)]
[mlir][ODS] Get rid of limitations in rewriters generator

Do not limit the number of arguments in rewriter pattern.

Introduce separate `FmtStrVecObject` class to handle
format of variadic `std::string` array.

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D97839

3 years ago[X86][NFC] Pre-commit test case for the fix of ldtilecfg insertion.
Wang, Pengfei [Thu, 18 Mar 2021 09:01:06 +0000 (17:01 +0800)]
[X86][NFC] Pre-commit test case for the fix of ldtilecfg insertion.

3 years ago[X86][AMX][NFC] Give correct Passname for Tile Register Pre-configure
Bing1 Yu [Thu, 18 Mar 2021 09:07:49 +0000 (17:07 +0800)]
[X86][AMX][NFC] Give correct Passname for Tile Register Pre-configure

3 years ago[-Wcalled-once-parameter] Harden analysis in terms of block use
Valeriy Savchenko [Mon, 15 Mar 2021 19:00:07 +0000 (22:00 +0300)]
[-Wcalled-once-parameter] Harden analysis in terms of block use

This patch introduces a very simple inter-procedural analysis
between blocks and enclosing functions.

We always analyze blocks first (analysis is done as part of semantic
analysis that goes side-by-side with the parsing process), and at the
moment of reporting we don't know how that block will be actually
used.

This patch introduces new logic delaying reports of the "never called"
warnings on blocks.  If we are not sure that the block will be called
exactly once, we shouldn't warn our users about that.  Double calls,
however, don't require such delays.  While analyzing the enclosing
function, we can actually decide what we should do with those
warnings.

Additionally, as a side effect, we can be more confident about blocks
in such context and can treat them not as escapes, but as direct
calls.

rdar://74090107

Differential Revision: https://reviews.llvm.org/D98688

3 years ago[RISCV] Support bitcasts of fixed-length mask vectors
Fraser Cormack [Wed, 17 Mar 2021 12:33:59 +0000 (12:33 +0000)]
[RISCV] Support bitcasts of fixed-length mask vectors

Without this patch, bitcasts of fixed-length mask vectors would go
through the stack.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D98779

3 years ago[X86] Fix compile time regression of D93594.
Luo, Yuanke [Wed, 17 Mar 2021 11:17:18 +0000 (19:17 +0800)]
[X86] Fix compile time regression of D93594.

D93594 depend on the dominate tree and loop information. It increased
the compile time when build with -O0. However this is just to amend the
dominate tree and loop information, so that it is unnecessary to
re-analyze them again. Given the dominate tree of loop information are
absent in this pass, we can avoid amending them.

Differential Revision: https://reviews.llvm.org/D98773

3 years ago[AArch64] Rewrite (add, csel) to cinc
Sjoerd Meijer [Tue, 16 Mar 2021 11:53:43 +0000 (11:53 +0000)]
[AArch64] Rewrite (add, csel) to cinc

Don't rewrite an add instruction with 2 SET_CC operands into a csel
instruction. The total instruction sequence uses an extra instruction and
register. Preventing this allows us to match a `(add, csel)` pattern and
rewrite this into a `cinc`.

Differential Revision: https://reviews.llvm.org/D98704

3 years ago[CMake] Use compiler-rt location instead of resource directory to find clang-cls...
Markus Böck [Thu, 18 Mar 2021 08:24:49 +0000 (09:24 +0100)]
[CMake] Use compiler-rt location instead of resource directory to find clang-cls runtime directory

The current cmake script attempts to add the path containing clangs various runtime systems by getting the resource directory and then appending the hardcoded value /lib/windows to it. This works for a normal clang-cl build but fails for a build of clang using LLVM_ENABLE_PER_TARGET_RUNTIME_DIR, such as the builds from llvm/runtimes.

This patch instead uses -print-libgcc-file-name in conjunction with --rtlib=compiler-rt, and instead adds the containing directory as library path.

For non per-target runtime directory builds, such as the release builds, there is no change. Even if the builtins library were to be deleted or moved it would output the same path as before.
For per-target runtime builds that also have the builtins library, this now finds the correct directory containing all of clang runtime libraries.

Only case still not handled by this change, is if a per-target runtime directory build is used, but the builtins library was not built.
I believe that is the best we can do for now however, without modifying clang.

Differential Revision: https://reviews.llvm.org/D98786

3 years agoRevert "[NFC] Minor cleanup to use default setting of getLastArg()"
Maxim Kuvyrkov [Thu, 18 Mar 2021 08:05:14 +0000 (08:05 +0000)]
Revert "[NFC] Minor cleanup to use default setting of getLastArg()"

The patch was wrong.  We use "const Arg *A" at the end of
GetLinkerPath, so can't remove it.

This reverts commit 6802fdf8871f69d52b06d0a2b7f62f3af8292690.

3 years ago[MLIR] Canonicalize broadcast operations on single shapes
Frederik Gossen [Thu, 18 Mar 2021 07:58:59 +0000 (08:58 +0100)]
[MLIR] Canonicalize broadcast operations on single shapes

This covers cases that are not folded away because the extent tensor type
becomes more concrete in the process.

Differential Revision: https://reviews.llvm.org/D98782

3 years ago[NFC] Minor cleanup to use default setting of getLastArg()
Maxim Kuvyrkov [Thu, 18 Mar 2021 07:47:16 +0000 (07:47 +0000)]
[NFC] Minor cleanup to use default setting of getLastArg()

Noticed this while I was looking at linker defaults.

Reviewed By: asl

Differential Revision: https://reviews.llvm.org/D98494

3 years ago[WoA][MSVC] Use default linker setting in MSVC-compatible driver
Maxim Kuvyrkov [Thu, 18 Mar 2021 07:42:41 +0000 (07:42 +0000)]
[WoA][MSVC] Use default linker setting in MSVC-compatible driver

At the moment "link.exe" is hard-coded as default linker in MSVC.cpp,
so there's no way to use LLD as default linker for MSVC driver.

This patch adds checking of CLANG_DEFAULT_LINKER to MSVC.cpp.

Reviewed By: asl

Differential Revision: https://reviews.llvm.org/D98493

3 years ago[JITLink] Reformat an enum.
Lang Hames [Thu, 18 Mar 2021 04:43:53 +0000 (21:43 -0700)]
[JITLink] Reformat an enum.

3 years ago[JITLink] Improve out-of-range error messages.
Lang Hames [Thu, 18 Mar 2021 04:19:13 +0000 (21:19 -0700)]
[JITLink] Improve out-of-range error messages.

Switches all backends to use the makeTargetOutOfRangeError function from
JITLink.h.

3 years ago[analyzer] Introduce common bug category "Unused code".
Artem Dergachev [Thu, 18 Mar 2021 03:34:34 +0000 (20:34 -0700)]
[analyzer] Introduce common bug category "Unused code".

This category is generic enough to hold a variety of checkers.
Currently it contains the Dead Stores checker and an alpha unreachable
code checker.

Differential Revision: https://reviews.llvm.org/D98741

3 years ago[RISCV][Clang] Add RVV vle/vse intrinsic functions.
Zakk Chen [Wed, 17 Mar 2021 14:56:55 +0000 (07:56 -0700)]
[RISCV][Clang] Add RVV vle/vse intrinsic functions.

Add new field PermuteOperands to mapping different operand order between
C/C++ API and clang builtin.

Reviewed By: craig.topper, rogfer01

Authored-by: Roger Ferrer Ibanez <rofirrim@gmail.com>
Co-Authored-by: Hsiangkai Wang <kai.wang@sifive.com>
Co-Authored-by: Zakk Chen <zakk.chen@sifive.com>
Differential Revision: https://reviews.llvm.org/D98388

3 years ago[Clang][RISCV] Add rvv vsetvl and vsetvlmax intrinsic functions.
Zakk Chen [Thu, 25 Feb 2021 08:15:14 +0000 (00:15 -0800)]
[Clang][RISCV] Add rvv vsetvl and vsetvlmax intrinsic functions.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D96843

3 years ago[RISCV] Fix isel pattern of masked vmslt[u]
ShihPo Hung [Thu, 18 Mar 2021 02:08:46 +0000 (19:08 -0700)]
[RISCV] Fix isel pattern of masked vmslt[u]

This patch changes the operand order of masked vmslt[u]
from (mask, rs1, scalar, maskedoff, vl)
to (maskedoff, rs1, scalar, mask, vl).

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D98839

3 years ago[sanitizer] Grow buffer in SharedPrintfCodeNoBuffer
Vitaly Buka [Thu, 18 Mar 2021 03:14:04 +0000 (20:14 -0700)]
[sanitizer] Grow buffer in SharedPrintfCodeNoBuffer

3 years ago[ObjectYAML] Handle Hexagon V68
Krzysztof Parzyszek [Thu, 18 Mar 2021 02:37:40 +0000 (21:37 -0500)]
[ObjectYAML] Handle Hexagon V68

3 years ago[NFC][sanitizer] Remove unneeded "explicit"
Vitaly Buka [Thu, 18 Mar 2021 02:40:45 +0000 (19:40 -0700)]
[NFC][sanitizer] Remove unneeded "explicit"

3 years ago[Hexagon] Improve stack address base reuse for HVX spills
Krzysztof Parzyszek [Thu, 18 Mar 2021 02:14:35 +0000 (21:14 -0500)]
[Hexagon] Improve stack address base reuse for HVX spills

The offset in HVX loads/stores is only 4 bits long, so often an
extra register is needed to hold the address. Minimize the number
of such registers by "standardizing" the base addresses and reusing
preexisting base registers when replacing frame indices.

3 years ago[Hexagon] Add more patterns for HVX loads and stores
Krzysztof Parzyszek [Thu, 18 Mar 2021 01:57:37 +0000 (20:57 -0500)]
[Hexagon] Add more patterns for HVX loads and stores

In particular, add patterns for loads/stores to the stack
(with a frame index as address).

3 years ago[NFC] make XCOFF dwarf dump test run only on PowerPC target.
Chen Zheng [Thu, 18 Mar 2021 01:58:09 +0000 (21:58 -0400)]
[NFC] make XCOFF dwarf dump test run only on PowerPC target.

3 years ago[XCOFF][llvm-dwarfdump] llvm-dwarfdump support for XCOFF
Chen Zheng [Tue, 16 Mar 2021 06:08:57 +0000 (02:08 -0400)]
[XCOFF][llvm-dwarfdump] llvm-dwarfdump support for XCOFF

Author: hubert.reinterpretcast, shchenz

Reviewed By: jasonliu, echristo

Differential Revision: https://reviews.llvm.org/D97186

3 years ago[GlobalISel] Don't DCE LIFETIME_START/LIFETIME_END markers.
Amara Emerson [Wed, 17 Mar 2021 18:34:56 +0000 (11:34 -0700)]
[GlobalISel] Don't DCE LIFETIME_START/LIFETIME_END markers.

These are pseudos without any users, so DCE was killing them in the combiner.

Marking them as having side effects doesn't seem quite right since they don't.

Gives a nice 0.3% geomean size win on CTMark -Os.

Differential Revision: https://reviews.llvm.org/D98811

3 years ago[AMDGPU] Avoid unnecessary graph visits during WQM marking
Carl Ritson [Thu, 18 Mar 2021 00:31:39 +0000 (09:31 +0900)]
[AMDGPU] Avoid unnecessary graph visits during WQM marking

Avoid revisiting nodes with the same set of defined lanes by
using a unified visited set which integrates lanes into the key.
This retains the intent of the original code by still revisiting
a subgraph if a different set of lanes is defined and hence
marking might progress differently.

Note: default size of the visited set has been confirmed to
cover >99% of invocations in large array of test shaders.

Reviewed By: piotr

Differential Revision: https://reviews.llvm.org/D98772

3 years ago[mlir][Toy] Tidy up the first half of Chapter 2.
River Riddle [Thu, 18 Mar 2021 00:36:42 +0000 (17:36 -0700)]
[mlir][Toy] Tidy up the first half of Chapter 2.

This performs a few rewordings, expands on a few parts, etc.

3 years ago[mlir][Toy] Update the tutorial to use tablegen for dialect declarations
River Riddle [Thu, 18 Mar 2021 00:36:31 +0000 (17:36 -0700)]
[mlir][Toy] Update the tutorial to use tablegen for dialect declarations

This was missed when the feature was originally added.

Differential Revision: https://reviews.llvm.org/D87060

3 years ago[mlir][tosa] Add tosa.slice to std.subtensor lowering
Rob Suderman [Wed, 17 Mar 2021 22:53:18 +0000 (15:53 -0700)]
[mlir][tosa] Add tosa.slice to std.subtensor lowering

Lowering to subtensor is added for tosa.slice operator.

Differential Revision: https://reviews.llvm.org/D98825

3 years agoRevert "[CodeGenModule] Set dso_local for Mach-O GlobalValue"
Alex Lorenz [Thu, 11 Mar 2021 22:54:47 +0000 (14:54 -0800)]
Revert "[CodeGenModule] Set dso_local for Mach-O GlobalValue"

This reverts commit 809a1e0ffd7af40ee27270ff8ba2ffc927330e71.

Mach-O doesn't support dso_local and this change broke XNU because of the use of dso_local.

Differential Revision: https://reviews.llvm.org/D98458

3 years agoPR49619: Remove delayed call to noteFailed.
Richard Smith [Thu, 18 Mar 2021 00:20:46 +0000 (17:20 -0700)]
PR49619: Remove delayed call to noteFailed.

This would assert if we hit the evaluation step limit between starting
to delay the call and finishing. In any case, delaying the call was
largely pointless as it doesn't really matter when we mark the
evaluation as having had side effects.

3 years ago[sanitizer] Remove max_len parameter from InternalScopedString
Vitaly Buka [Tue, 16 Mar 2021 23:33:04 +0000 (16:33 -0700)]
[sanitizer] Remove max_len parameter from InternalScopedString

InternalScopedString uses InternalMmapVector internally
so it can be resized dynamically as needed.

Reviewed By: eugenis

Differential Revision: https://reviews.llvm.org/D98751

3 years ago[FileCheck] Fix redundant diagnostics due to numeric errors
Joel E. Denny [Wed, 17 Mar 2021 18:17:41 +0000 (14:17 -0400)]
[FileCheck] Fix redundant diagnostics due to numeric errors

Fixed substitution printing not to produce an empty diagnostic for
errors handled elsewhere.

Reviewed By: thopre

Differential Revision: https://reviews.llvm.org/D98088

3 years ago[FileCheck] Fix numeric error propagation
Joel E. Denny [Wed, 17 Mar 2021 18:13:57 +0000 (14:13 -0400)]
[FileCheck] Fix numeric error propagation

A more general name might be match-time error propagation.  That is,
it's conceivable we'll one day have non-numeric errors that require
the handling fixed by this patch.

Without this patch, FileCheck behaves as follows:

```
$ cat check
CHECK-NOT: [[#0x8000000000000000+0x8000000000000000]]

$ FileCheck -vv -dump-input=never check < input
check:1:54: remark: implicit EOF: expected string found in input
CHECK-NOT: [[#0x8000000000000000+0x8000000000000000]]
                                                     ^
<stdin>:2:1: note: found here

^
check:1:15: error: unable to substitute variable or numeric expression: overflow error
CHECK-NOT: [[#0x8000000000000000+0x8000000000000000]]
              ^
$ echo $?
0
```

Notice that the exit status is 0 even though there's an error.
Moreover, FileCheck doesn't print the error diagnostic unless both
`-dump-input=never` and `-vv` are specified.

The same problem occurs when `CHECK-NOT` does have a match but a
capture fails due to overflow: exit status is 0, and no diagnostic is
printed unless both `-dump-input=never` and `-vv` are specified.  The
usefulness of capturing from `CHECK-NOT` is questionable, but this
case should certainly produce an error.

With this patch, FileCheck always includes the error diagnostic and
has non-zero exit status for the above examples.  It's conceivable
that this change will cause some existing tests to fail, but my
assumption is that they should fail.  Moreover, with nearly every
project enabled, this patch didn't produce additional `check-all`
failures for me.

This patch also extends input dumps to include such numeric error
diagnostics for both expected and excluded patterns.

As noted in fixmes in some of the tests added by this patch, this
patch worsens an existing issue with redundant diagnostics.  I'll fix
that bug in a subsequent patch.

Reviewed By: thopre, jhenderson

Differential Revision: https://reviews.llvm.org/D98086

3 years agoPR49585: Emit the jump destination for a for loop 'continue' from within the scope...
Richard Smith [Wed, 17 Mar 2021 21:00:03 +0000 (14:00 -0700)]
PR49585: Emit the jump destination for a for loop 'continue' from within the scope of the condition variable.

The condition variable is in scope in the loop increment, so we need to
emit the jump destination from wthin the scope of the condition
variable.

For GCC compatibility (and compatibility with real-world 'FOR_EACH'
macros), 'continue' is permitted in a statement expression within the
condition of a for loop, though, so there are two cases here:

* If the for loop has no condition variable, we can emit the jump
  destination before emitting the condition.

* If the for loop has a condition variable, we must defer emitting the
  jump destination until after emitting the variable. We diagnose a
  'continue' appearing in the initializer of the condition variable,
  because it would jump past the initializer into the scope of that
  variable.

Reviewed By: rjmccall

Differential Revision: https://reviews.llvm.org/D98816

3 years ago[OPENMP51]Initial support for the use clause.
Mike Rice [Wed, 17 Mar 2021 20:04:08 +0000 (13:04 -0700)]
[OPENMP51]Initial support for the use clause.

Added basic parsing/sema/serialization support for the 'use' clause.

Differential Revision: https://reviews.llvm.org/D98815

3 years agoRevert "[NewPM] Verify LoopAnalysisResults after a loop pass"
Arthur Eubanks [Wed, 17 Mar 2021 22:22:35 +0000 (15:22 -0700)]
Revert "[NewPM] Verify LoopAnalysisResults after a loop pass"

This reverts commit 6db3ab2903f42712f44000afb5aa467efbd25f35.

Causing too large of compile time regression.

3 years ago[lld-macho][NFC] Minor refactor of Writer::run()
Greg McGary [Sun, 14 Mar 2021 22:35:27 +0000 (15:35 -0700)]
[lld-macho][NFC] Minor refactor of Writer::run()

Move some functions closer to their uses. Move detailed address-assignment logic out of the otherwise abstract `Writer::run()`. This prepares the ground for a diff to implement branch range extension thunks.

* `SyntheticSections.cpp`
 ** move `needsBinding()` and `prepareBranchTarget()` into `Writer.cpp`
 ** move `addNonLazyBindingEntries()` adjacent to its use.

* `Writer.cpp`
 ** move address-assignment logic from `Writer::run()` into new function `Writer::assignAddresses()`
 ** move `needsBinding()` and `prepareBranchTarget()` from `SyntheticSections.cpp`

* `Target.h`
** remove orphaned decls of `prepareSymbolRelocation()` and `validateRelocationInfo()` which were moved to other files in earlier diffs.

Differential Revision: https://reviews.llvm.org/D98795

3 years ago[AArch64][GlobalISel] Fall back if disabling neon/fp in the translator.
Amara Emerson [Tue, 16 Mar 2021 18:56:32 +0000 (11:56 -0700)]
[AArch64][GlobalISel] Fall back if disabling neon/fp in the translator.

The previous technique relied on early-exiting the legalizer predicate
initialization, leaving an empty rule table. That causes a fallback
for most instructions, but some have legacy rules defined like G_ZEXT
which can try continue, but then crash.

We should fall back earlier, in the translator, to avoid this issue.

Differential Revision: https://reviews.llvm.org/D98730

3 years ago[Object][MachO] Handle end iterator in getSymbolType()
Steven Wu [Wed, 17 Mar 2021 22:05:51 +0000 (15:05 -0700)]
[Object][MachO] Handle end iterator in getSymbolType()

Fix a bug in MachOObjectFile::getSymbolType() that it is not checking if
the iterator is end() before deference the iterator. Instead, return
`Other` type, which aligns with the behavior of `llvm-nm`.

rdar://75291638

Reviewed By: davide, ab

Differential Revision: https://reviews.llvm.org/D98739

3 years ago[ASTMatchers][NFC] Use move semantics when passing matchers around.
Nathan James [Wed, 17 Mar 2021 22:03:07 +0000 (22:03 +0000)]
[ASTMatchers][NFC] Use move semantics when passing matchers around.

Changing matchers to use non-const members and adding r-value overloads of matcher conversions enables move optimisations.
I don't have performance figures but I can say this knocked 120k from the clang-tidy binary(86k was from the .text section) on a Release with assertions build(x86_64-unknown-linux-gnu).

Reviewed By: aaron.ballman

Differential Revision: https://reviews.llvm.org/D98792

3 years ago[ARM] Add VREV MVE shuffle costs
David Green [Wed, 17 Mar 2021 21:21:43 +0000 (21:21 +0000)]
[ARM] Add VREV MVE shuffle costs

This uses the shuffle mask cost from D98206 to give a better cost of MVE
VREV instructions. This helps especially in VectorCombine where the cost
of shuffles is used to reorder bitcasts, which this helps keep the phase
ordering test for fp16 reductions producing optimal code. The isVREVMask
has been moved to a header file to allow it to be used across target
transform and isel lowering.

Differential Revision: https://reviews.llvm.org/D98210

3 years ago[mlir][IR] Support parsing hex float values in the DialectSymbolParser
River Riddle [Wed, 17 Mar 2021 20:09:53 +0000 (13:09 -0700)]
[mlir][IR] Support parsing hex float values in the DialectSymbolParser

This has been a TODO for a while, and prevents breakages for attributes/types that contain floats that can't roundtrip outside of the hex format.

Differential Revision: https://reviews.llvm.org/D98808

3 years ago[NewPM] Verify LoopAnalysisResults after a loop pass
Arthur Eubanks [Wed, 17 Mar 2021 19:21:59 +0000 (12:21 -0700)]
[NewPM] Verify LoopAnalysisResults after a loop pass

All loop passes should preserve all analyses in LoopAnalysisResults. Add
checks for those.

Note that due to PR44815, we don't check LAR's ScalarEvolution.
Apparently calling SE.verify() can change its results.

Reviewed By: asbirlea

Differential Revision: https://reviews.llvm.org/D98805

3 years ago[M68k] Forward declare getMCInstrBeads in one place
Ricky Taylor [Wed, 17 Mar 2021 20:30:39 +0000 (13:30 -0700)]
[M68k] Forward declare getMCInstrBeads in one place

At the moment `getMCInstrBeads` is forward-declared in a few places,
bring this together into a single header file.

This was done as part of the disassembler work, since the disassembler
would otherwise add one more forward declaration.

Differential Revision: https://reviews.llvm.org/D98533

3 years ago[M68k] Use fixed asm string for MxPseudo instructions
Ricky Taylor [Wed, 17 Mar 2021 20:29:02 +0000 (13:29 -0700)]
[M68k] Use fixed asm string for MxPseudo instructions

This is required because empty strings are not allowed when generating
the assembly parser tables.

Differential Revision: https://reviews.llvm.org/D98532

3 years ago[compiler-rt] -fsanitize=cfi is not supported on Darwin
Jon Roelofs [Wed, 17 Mar 2021 17:50:59 +0000 (10:50 -0700)]
[compiler-rt] -fsanitize=cfi is not supported on Darwin

This was responsible for:

Failed Tests (2):
  cfi-devirt-x86_64 :: mfcall.cpp
  cfi-standalone-x86_64 :: mfcall.cpp

3 years ago[NFC][AArch64] Add codegen tests for various csinc-cmp sequences.
Pavel Iliin [Wed, 17 Mar 2021 01:15:00 +0000 (01:15 +0000)]
[NFC][AArch64] Add codegen tests for various csinc-cmp sequences.

3 years ago[X86][SSE] Add SSE2/SSE42 test coverage to urem combine tests
Simon Pilgrim [Wed, 17 Mar 2021 19:57:53 +0000 (19:57 +0000)]
[X86][SSE] Add SSE2/SSE42 test coverage to urem combine tests

Noticed when reviewing D88785

3 years ago[lld-link] emit an error when writing a PDB > 4 GiB
Nico Weber [Wed, 17 Mar 2021 15:56:13 +0000 (11:56 -0400)]
[lld-link] emit an error when writing a PDB > 4 GiB

Maybe there's a way to make them work, but until I've investigated
if tools can consume large PDBs, erroring out is better than slowly
and silently consuming all available ram due to internal invariants
being violated.

(Patch to make writing larger files work at
https://bugs.chromium.org/p/chromium/issues/detail?id=1179085#c25
but I haven't had time to check if windbg & co can consume these
large PDBs. llvm-pdbutil can't, but we can fix that one at least :) )

Differential Revision: https://reviews.llvm.org/D98788

3 years ago[LCSSA] Extract a utility for deciding if a new use requires a new lcssa phi [NFC]
Philip Reames [Wed, 17 Mar 2021 19:12:00 +0000 (12:12 -0700)]
[LCSSA] Extract a utility for deciding if a new use requires a new lcssa phi [NFC]

(Triggered by a review comment on D98728, but otherwise unrelated.)

3 years ago[Test] Fix undef var in attr-speculative-load-hardening.c
Thomas Preud'homme [Tue, 15 Dec 2020 23:05:45 +0000 (23:05 +0000)]
[Test] Fix undef var in attr-speculative-load-hardening.c

Fix use of undefined variable in CHECK-NOT directive in clang test
CodeGen/attr-speculative-load-hardening.c.

Reviewed By: kristof.beyls

Differential Revision: https://reviews.llvm.org/D93347

3 years ago[RISCV] Use getTargetExtractSubreg and getTargetInsertSubreg to simplify some code...
Craig Topper [Wed, 17 Mar 2021 19:09:48 +0000 (12:09 -0700)]
[RISCV] Use getTargetExtractSubreg and getTargetInsertSubreg to simplify some code. NFCI

3 years ago[lldb/Docs] Update docs with new buildbot URLs
Jan Kratochvil [Wed, 17 Mar 2021 19:09:03 +0000 (20:09 +0100)]
[lldb/Docs] Update docs with new buildbot URLs

3 years ago[lldb] [test] Skip vCont tests on Windows
Michał Górny [Wed, 17 Mar 2021 18:22:58 +0000 (19:22 +0100)]
[lldb] [test] Skip vCont tests on Windows

3 years ago[LICM] Fix a crash when sinking instructions w/token operands
Philip Reames [Wed, 17 Mar 2021 17:57:33 +0000 (10:57 -0700)]
[LICM] Fix a crash when sinking instructions w/token operands

It is not legal to form a phi node with token type. The generic LCSSA construction code handles this correctly - by not forming LCSSA for such cases - but the adhoc fixup implementation in LICM did not.

This was noticed in the context of PR49607, but can be demonstrated on ToT with the tweaked test case. This is not specific to gc.relocate btw, it also applies to usage of the preallocated family of intrinsics as well.

Differential Revision: https://reviews.llvm.org/D98728

3 years ago[RISCV] Update RVV shift intrinsic tests to use XLEN bit as shift amount.
Zakk Chen [Thu, 11 Mar 2021 16:01:04 +0000 (08:01 -0800)]
[RISCV] Update RVV shift intrinsic tests to use XLEN bit as shift amount.

Fix the unexpected of using op1's element type as shift amount type.

Reviewed By: frasercrmck

Differential Revision: https://reviews.llvm.org/D98501

3 years ago[TTI] Add a Mask to getShuffleCost
David Green [Wed, 17 Mar 2021 17:46:26 +0000 (17:46 +0000)]
[TTI] Add a Mask to getShuffleCost

This adds an Mask ArrayRef to getShuffleCost, so that if an exact mask
can be provided a more accurate cost can be provided by the backend.
For example VREV costs could be returned by the ARM backend. This should
be an NFC until then, laying the groundwork for that to be added.

Differential Revision: https://reviews.llvm.org/D98206

3 years ago[RISCV] Support masked load/store for fixed vectors.
Craig Topper [Wed, 17 Mar 2021 17:13:00 +0000 (10:13 -0700)]
[RISCV] Support masked load/store for fixed vectors.

Reviewed By: frasercrmck

Differential Revision: https://reviews.llvm.org/D98561

3 years ago[mlir][cpu-runner] register all llvm ir dialects
Aart Bik [Tue, 16 Mar 2021 18:05:09 +0000 (11:05 -0700)]
[mlir][cpu-runner] register all llvm ir dialects

This fixes broken JIT functionality on emulator platforms.
With Alex' recent movement towards squashing llvm ir dialects
into target specific dialects, we now must ensure these dialects
are registered to the cpu runner to ensure JIT can lower this
to proper LLVM IR before handing this off to the backend.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D98727

3 years ago[mlir][amx] regression test for tile-muli (all zero/sign-extension combinations)
Aart Bik [Tue, 16 Mar 2021 20:37:34 +0000 (13:37 -0700)]
[mlir][amx] regression test for tile-muli (all zero/sign-extension combinations)

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D98742

3 years agoReapply "[DebugInfo] Handle multiple variable location operands in IR"
Stephen Tozer [Wed, 17 Mar 2021 15:04:27 +0000 (15:04 +0000)]
Reapply "[DebugInfo] Handle multiple variable location operands in IR"

Fixed section of code that iterated through a SmallDenseMap and added
instructions in each iteration, causing non-deterministic code; replaced
SmallDenseMap with MapVector to prevent non-determinism.

This reverts commit 01ac6d1587e8613ba4278786e8341f8b492ac941.

3 years ago[OPENMP51]Initial support for the interop directive.
Mike Rice [Mon, 15 Mar 2021 20:09:46 +0000 (13:09 -0700)]
[OPENMP51]Initial support for the interop directive.

Added basic parsing/sema/serialization support for interop directive.
Support for the 'init' clause.

Differential Revision: https://reviews.llvm.org/D98558

3 years ago[CGSCC] Print CG node itself instead of its address
Bardia Mahjour [Wed, 17 Mar 2021 16:36:55 +0000 (12:36 -0400)]
[CGSCC] Print CG node itself instead of its address

Fix the debug output from cgscc

3 years ago[ELF] Special case --shuffle-sections=-1 to reverse input sections
Fangrui Song [Wed, 17 Mar 2021 16:32:44 +0000 (09:32 -0700)]
[ELF] Special case --shuffle-sections=-1 to reverse input sections

If the number of sections changes, which is common for re-links after
incremental updates, the section order may change drastically.

Special case -1 to reverse input sections. This is a stable transform.
The section order is more resilient to incremental updates.  Usually the
code issue (e.g. Static Initialization Order Fiasco, assuming pointer
comparison result of two unrelated objects) is due to the relative order
between two problematic input files A and B.  Checking the regular order
and the reversed order is sufficient.

Differential Revision: https://reviews.llvm.org/D98445

3 years ago[lldb] [test] Fix TestGdbRemote_vContThreads.py logic
Michał Górny [Tue, 16 Mar 2021 23:04:06 +0000 (00:04 +0100)]
[lldb] [test] Fix TestGdbRemote_vContThreads.py logic

The TestGdbRemote_vContThreads.py were introduced to test NetBSD process
plugin's capability of sending per-thread and per-process signals.
However, at some point the tests started failing.  From retrospective,
it is possible that they were relying on some bug in the plugin's
original signal handling.

Fix the tests not to expect the process to terminate after receiving
the signals.  Instead, scan for output indicating that the signals were
received and match thread IDs in it.  Enable 'signal to all threads'
test everywhere as it works fine on Linux.  Add a new test for vCont
packet without specific thread IDs.  Introduce a helper function
to cover the common part of tests.

While this does not fix all the problems on NetBSD, it enables a subset
of the tests on other systems.  I am planning to add more tests
to the group while implementing multiprocess extension for vCont.

Differential Revision: https://reviews.llvm.org/D98749

3 years ago[ms] [llvm-ml] Allow the /Zs parameter as a synonym for -filetype=null
Eric Astor [Wed, 17 Mar 2021 16:10:11 +0000 (12:10 -0400)]
[ms] [llvm-ml] Allow the /Zs parameter as a synonym for -filetype=null

For ml.exe, /Zs implies a syntax check with no output files.

Reviewed By: thakis

Differential Revision: https://reviews.llvm.org/D90061

3 years ago[lldb] Correct unsigned decimal argument check in memory write
David Spickett [Wed, 17 Mar 2021 15:45:37 +0000 (15:45 +0000)]
[lldb] Correct unsigned decimal argument check in memory write

getAsInteger returns false when it succeeds.

Before:
(lldb) memory write 0x00007ffff7dd3000 99 -f "unsigned decimal"
error: '99' is not a valid unsigned decimal string value.

After:
(lldb) memory write 0x00007ffff7dd3000 99 -f "unsigned decimal"
(lldb) memory read 0x00007ffff7dd3000 0x00007ffff7dd3001
0x7ffff7dd3000: 63                                               c

3 years ago[LoopVectorize] Refine hasIrregularType predicate
LemonBoy [Wed, 17 Mar 2021 15:59:55 +0000 (16:59 +0100)]
[LoopVectorize] Refine hasIrregularType predicate

The `hasIrregularType` predicate checks whether an array of N values of type Ty is "bitcast-compatible" with a <N x Ty> vector.
The previous check returned invalid results in some cases where there's some padding between the array elements: eg. a 4-element array of u7 values is considered as compatible with <4 x u7>, even though the vector is only loading/storing 28 bits instead of 32.

The problem causes LLVM to generate incorrect code for some targets: for AArch64 the vector loads/stores are lowered in terms of ubfx/bfi, effectively losing the top (N * padding bits).

Reviewed By: lebedev.ri

Differential Revision: https://reviews.llvm.org/D97465

3 years ago[Sanitizer] Fix debug builds of sanitizer_stacktrace_test.cpp
Luís Marques [Wed, 17 Mar 2021 15:57:00 +0000 (15:57 +0000)]
[Sanitizer] Fix debug builds of sanitizer_stacktrace_test.cpp

An implementation of `__sanitizer::BufferedStackTrace::UnwindImpl` is
provided per sanitizer, but there isn't one for sanitizer-common. In
non-optimized builds of the sanitizer-common tests that becomes a problem:
the test `sanitizer_stacktrace_test.cpp` won't have a reference to that
method optimized away, causing linking errors. This patch provides a dummy
implementation, which fixes those builds.

Differential Revision: https://reviews.llvm.org/D96956

3 years ago[ARM] Use lrdsb for more thumb1 loads.
David Green [Wed, 17 Mar 2021 15:29:02 +0000 (15:29 +0000)]
[ARM] Use lrdsb for more thumb1 loads.

Given a sextload i16, we can usually generate "ldrsh [rn. rm]". If we
don't naturally have a rn, rm addressing mode, we can either generate
"ldrh [rn, #0]; sxth" or "mov rm, #0; ldrsh [rn. rm]".

We currently generate the first, always creating a sxth. They are both
the same number of instructions, but if we generate the second then the
mov #0 will likely be CSE'd or pulled out of a loop, etc.

This adjusts the ISel patterns to do that, creating a mov instead of a
sxth.

Differential Revision: https://reviews.llvm.org/D98693

3 years agoAdd arm_neon.sdot operation
Ahmed Taei [Sat, 6 Mar 2021 04:54:01 +0000 (20:54 -0800)]
Add arm_neon.sdot operation

Differential Revision: https://reviews.llvm.org/D98198

3 years ago[DAG] TargetLowering::isBinOp() - add ISD::SSUBSAT/USUBSAT
Simon Pilgrim [Wed, 17 Mar 2021 14:26:25 +0000 (14:26 +0000)]
[DAG] TargetLowering::isBinOp() - add ISD::SSUBSAT/USUBSAT

Add to the generic non-commutative binop list.

3 years ago[RGT] RPCUtilsTest, replace un-executed EXPECT with unreachable
Paul Robinson [Fri, 12 Mar 2021 16:47:48 +0000 (08:47 -0800)]
[RGT] RPCUtilsTest, replace un-executed EXPECT with unreachable

Unreachable code should be self-documented as unreachable.

Found by the Rotten Green Tests project.

Differential Revision: https://reviews.llvm.org/D98518

3 years ago[llvm-objcopy][NFC] Move ownership keeping code into restoreStatOnFile().
Alexey Lapshin [Fri, 12 Mar 2021 13:31:35 +0000 (16:31 +0300)]
[llvm-objcopy][NFC] Move ownership keeping code into restoreStatOnFile().

The D93881 added functionality which preserve ownership for output file
if llvm-objcopy is called under root. That code was added into the place
where output file is created. The llvm-objcopy already has a function which
sets/restores rights/permissions for the output file.
That is the restoreStatOnFile() function. This patch moves code
(preserving ownershipping) into the restoreStatOnFile() function.

Differential Revision: https://reviews.llvm.org/D98511

3 years ago[clang] Disable LTO and LLD on SystemZ for stage3 builds
Timm Bäder [Tue, 9 Mar 2021 10:51:09 +0000 (11:51 +0100)]
[clang] Disable LTO and LLD on SystemZ for stage3 builds

LLD does not support SystemZ, so it doesn't make sense to use it for
boostrap/stage3 builds, and using LTO in these cases won't work.

Differential Revision: https://reviews.llvm.org/D89942

3 years ago[mlir][ODS] Support specialized Attribute class for Enums
Vladislav Vinogradov [Sat, 27 Feb 2021 12:21:00 +0000 (15:21 +0300)]
[mlir][ODS] Support specialized Attribute class for Enums

Add a feature to `EnumAttr` definition to generate
specialized Attribute class for the particular enumeration.

This class will inherit `StringAttr` or `IntegerAttr` and
will override `classof` and `getValue` methods.

With this class the enumeration predicate can be checked with simple
RTTI calls (`isa`, `dyn_cast`) and it will return the typed enumeration
directly instead of raw string/integer.

Based on the following discussion:
https://llvm.discourse.group/t/rfc-add-enum-attribute-decorator-class/2252

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D97836

3 years ago[OCaml] Handle nullptr in Llvm.global_initializer
Timotej Kapus [Wed, 17 Mar 2021 12:47:32 +0000 (12:47 +0000)]
[OCaml] Handle nullptr in Llvm.global_initializer

LLVMGetInitializer returns nullptr in case there is no initializer.
There is not much that can be done with nullptr in OCaml, not even
test if it is null. Also, there does not seem to be a C or OCaml API
to test if there is an initializer. So this diff changes
Llvm.global_initializer to return an option.

Reviewed By: whitequark

Differential Revision: https://reviews.llvm.org/D65195

3 years agoMove BaseOpWithOffsetSizesAndStrides to OpBase.td
Adrian Kuegel [Wed, 17 Mar 2021 12:43:18 +0000 (13:43 +0100)]
Move BaseOpWithOffsetSizesAndStrides to OpBase.td

It is used both by the Standard dialect and the MemRef dialect.

Differential Revision: https://reviews.llvm.org/D98777

3 years agoFixing a test case that was missed in c165a99a1b8861af87e0509a2e14debf2764804b
Aaron Ballman [Wed, 17 Mar 2021 12:45:28 +0000 (08:45 -0400)]
Fixing a test case that was missed in c165a99a1b8861af87e0509a2e14debf2764804b