[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
martinboehme wrote: > Updated to enforce llvm::SmallVector, per Martin's recommendation. PTAL. LGTM, but note that the PR title is now out of date. https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/martinboehme edited https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/Xazax-hun approved this pull request. https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
@@ -245,10 +245,10 @@ runDataflowAnalysis( /// iterations. /// - This limit is still low enough to keep runtimes acceptable (on typical /// machines) in cases where we hit the limit. -template -llvm::Expected> diagnoseFunction( +template +llvm::Expected diagnoseFunction( martinboehme wrote: ```suggestion template llvm::Expected> diagnoseFunction( ``` See also PR-level comment with rationale. https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/martinboehme edited https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/martinboehme requested changes to this pull request. https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/martinboehme review_requested https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
martinboehme wrote: I'm not convinced of the value of this change. Wouldn't it be simpler to keep the template arguments of `diagnoseFunction()` unchanged and instead change the return type to be `llvm::Expected>`? In other words, always return an `llvm::SmallVector` instead of a `std::vector`? I think `llvm::SmallVector` makes sense for both callsites that currently exist (and ones that we might add in the future). This patch currently lets `UncheckedOptionalAccessCheck::check()` continue to use `std::vector`, but I think `llvm::SmallVector` makes sense there as well, as in the common case, we expect only a small number of diagnostics or none at all. This seems like it would be true for any function that wants to call `diagnoseFunction()` in the future as well. https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/Xazax-hun approved this pull request. LGTM! https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/ymand review_requested https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/llvmbot labeled https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/ymand review_requested https://github.com/llvm/llvm-project/pull/66014 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [clang][dataflow] Change `diagnoseFunction` to take type of diagnostic list instead of diagnostic itself. (PR #66014)
https://github.com/ymand created
https://github.com/llvm/llvm-project/pull/66014:
The template is agnostic as to the type used by the list, as long as it is
compatible with `llvm::move` and `std::back_inserter`. In practice, we've
encountered analyses which use different types (`llvm::SmallVector` vs
`std::vector`), so it seems preferable to leave this open to the caller.
>From 6eda5a1fc6200027d6ae99dc5eff69aa88962c81 Mon Sep 17 00:00:00 2001
From: Yitzhak Mandelbaum
Date: Mon, 11 Sep 2023 21:34:00 +
Subject: [PATCH] [clang][dataflow] Change `diagnoseFunction` to take type of
diagnostic list instead of diagnostic itself.
The template is agnostic as to the type used by the list, as long as it is
compatible with `llvm::move` and `std::back_inserter`. In practice, we've
encountered analyses which use different types (`llvm::SmallVector` vs
`std::vector`), so it seems preferable to leave this open.
---
.../clang-tidy/bugprone/UncheckedOptionalAccessCheck.cpp | 4 ++--
.../clang/Analysis/FlowSensitive/DataflowAnalysis.h | 8
.../FlowSensitive/TypeErasedDataflowAnalysisTest.cpp | 4 ++--
3 files changed, 8 insertions(+), 8 deletions(-)
diff --git
a/clang-tools-extra/clang-tidy/bugprone/UncheckedOptionalAccessCheck.cpp
b/clang-tools-extra/clang-tidy/bugprone/UncheckedOptionalAccessCheck.cpp
index 183beb6bfb87d80..5811e2a4cd02266 100644
--- a/clang-tools-extra/clang-tidy/bugprone/UncheckedOptionalAccessCheck.cpp
+++ b/clang-tools-extra/clang-tidy/bugprone/UncheckedOptionalAccessCheck.cpp
@@ -56,8 +56,8 @@ void UncheckedOptionalAccessCheck::check(
// `diagnoseFunction` with config options.
if (llvm::Expected> Locs =
dataflow::diagnoseFunction(*FuncDecl,
*Result.Context,
- Diagnoser))
+ std::vector>(
+ *FuncDecl, *Result.Context, Diagnoser))
for (const SourceLocation : *Locs)
diag(Loc, "unchecked access to optional value");
else
diff --git a/clang/include/clang/Analysis/FlowSensitive/DataflowAnalysis.h
b/clang/include/clang/Analysis/FlowSensitive/DataflowAnalysis.h
index abd34f40922121e..fd2d7fce09073bf 100644
--- a/clang/include/clang/Analysis/FlowSensitive/DataflowAnalysis.h
+++ b/clang/include/clang/Analysis/FlowSensitive/DataflowAnalysis.h
@@ -245,10 +245,10 @@ runDataflowAnalysis(
/// iterations.
/// - This limit is still low enough to keep runtimes acceptable (on typical
/// machines) in cases where we hit the limit.
-template
-llvm::Expected> diagnoseFunction(
+template
+llvm::Expected diagnoseFunction(
const FunctionDecl , ASTContext ,
-llvm::function_ref(
+llvm::function_ref &)>
Diagnoser,
@@ -263,7 +263,7 @@ llvm::Expected> diagnoseFunction(
DataflowAnalysisContext AnalysisContext(std::move(OwnedSolver));
Environment Env(AnalysisContext, FuncDecl);
AnalysisT Analysis(ASTCtx);
- std::vector Diagnostics;
+ DiagnosticList Diagnostics;
if (llvm::Error Err =
runTypeErasedDataflowAnalysis(
*Context, Analysis, Env,
diff --git
a/clang/unittests/Analysis/FlowSensitive/TypeErasedDataflowAnalysisTest.cpp
b/clang/unittests/Analysis/FlowSensitive/TypeErasedDataflowAnalysisTest.cpp
index af34a2afd24685a..3a8253508016caa 100644
--- a/clang/unittests/Analysis/FlowSensitive/TypeErasedDataflowAnalysisTest.cpp
+++ b/clang/unittests/Analysis/FlowSensitive/TypeErasedDataflowAnalysisTest.cpp
@@ -98,12 +98,12 @@ TEST(DataflowAnalysisTest,
DiagnoseFunctionDiagnoserCalledOnEachElement) {
cast(findValueDecl(AST->getASTContext(), "target"));
auto Diagnoser = [](const CFGElement , ASTContext &,
const TransferStateForDiagnostics &) {
-std::vector Diagnostics(1);
+llvm::SmallVector Diagnostics(1);
llvm::raw_string_ostream OS(Diagnostics.front());
Elt.dumpToStream(OS);
return Diagnostics;
};
- auto Result = diagnoseFunction(
+ auto Result = diagnoseFunction>(
*Func, AST->getASTContext(), Diagnoser);
// `diagnoseFunction` provides no guarantees about the order in which
elements
// are visited, so we use `UnorderedElementsAre`.
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
