[clang] [analyzer][NFC] Add option assume-one-iteration (PR #125494)
Donát Nagy via cfe-commits
cfe-commits at lists.llvm.org
Mon Feb 3 05:18:22 PST 2025
https://github.com/NagyDonat created https://github.com/llvm/llvm-project/pull/125494
This commit adds the new analyzer option `assume-one-iteration`, which is `false` by default, but can be set to `true` to ensure that the analyzer always assumes (at least) one iteration in loops.
In some situations this "loop is skipped" execution path is an important corner case that may evade the notice of the developer and hide significant bugs -- however, there are also many situations where it's guaranteed that at least one iteration will happen (e.g. some data structure is always nonempty), but the analyzer cannot realize this and will produce false positives when it assumes that the loop is skipped.
This commit refactors some logic around the implementation of the new feature, but those changes are clearly NFC when the new analyzer option is in its default state.
-----
This commit is the new step of my [loop handling improvement plans](https://discourse.llvm.org/t/loop-handling-improvement-plans/80417). When I wrote this, I hoped (based on eyeballing older analysis run results) that this will be an important "must have" improvement of the analyzer results; however it turns out that the effects are more ambiguous.
Based on this, I no longer think that this option should be enabled by default, but I still think that this option could be useful for some users to reduce the amount of unwanted results.
I analyzed the usual set of open source projects to see the effects of enabling this option, and I started to evaluate the results, but I didn't finish this evaluation because I felt that even this partial evaluation is enough to justify merging this as an off-by-default feature.
| Project | Assuming one iteration | Baseline | Notes |
|---------|------------------------|----------| ----- |
| memcached | [0 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=memcached_1.6.8_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=memcached_1.6.8_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [1 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=memcached_1.6.8_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=memcached_1.6.8_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved) | assuming empty list -> calloc with zero size report |
| tmux | [0 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=tmux_2.6_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=tmux_2.6_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [7 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=tmux_2.6_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=tmux_2.6_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved) | all 7 are core.NonNullParamChecker reports that assume emptiness of custom data structures, e.g. [this](https://codechecker-demo.eastus.cloudapp.azure.com/Default/report-detail?run=tmux_2.6_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=tmux_2.6_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved&report-id=6875004&report-hash=db2dd8ecbd79514cf57950e9cdc116e0&report-filepath=tmux%2Fwindow-buffer.c) |
| curl | [0 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=curl_curl-7_66_0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=curl_curl-7_66_0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [0 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=curl_curl-7_66_0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=curl_curl-7_66_0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved) | |
| twin | [0 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [3 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved) | one [corner case](https://codechecker-demo.eastus.cloudapp.azure.com/Default/report-detail?run=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved&report-id=6876057&report-hash=ec682267df95619957c1f3658029ca16&report-filepath=twin%2Fserver%2Frcrun.c) that seems to be very unlikely but might be a true positive, one [false positive](https://codechecker-demo.eastus.cloudapp.azure.com/Default/report-detail?run=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved&report-id=6876104&report-hash=a71674c6b490eb0e63d8fd26c1fb7cb5&report-filepath=twin%2Flibs%2FlibTw%2FlibTw.c) assumes an empty string, one [unrelated false positive](https://codechecker-demo.eastus.cloudapp.azure.com/Default/report-detail?run=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=twin_v0.8.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved&report-id=6876107&report-hash=b922b94f0857d8c03c7a7f76d38e717d&report-filepath=twin%2Flibs%2FlibTw%2FlibTw.c) accidentally discarded
| vim | [5 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=vim_v8.2.1920_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=vim_v8.2.1920_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [18 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=vim_v8.2.1920_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=vim_v8.2.1920_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved) |
| openssl | [0 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=openssl_openssl-3.0.0-alpha7_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=openssl_openssl-3.0.0-alpha7_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [5 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=openssl_openssl-3.0.0-alpha7_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=openssl_openssl-3.0.0-alpha7_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| sqlite | [12 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=sqlite_version-3.33.0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=sqlite_version-3.33.0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [25 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=sqlite_version-3.33.0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=sqlite_version-3.33.0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| ffmpeg | [24 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=ffmpeg_n4.3.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=ffmpeg_n4.3.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [175 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=ffmpeg_n4.3.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=ffmpeg_n4.3.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| postgres | [2 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=postgres_REL_13_0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=postgres_REL_13_0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [49 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=postgres_REL_13_0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=postgres_REL_13_0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| tinyxml2 | [0 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=tinyxml2_8.0.0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=tinyxml2_8.0.0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [0 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=tinyxml2_8.0.0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=tinyxml2_8.0.0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| libwebm | [3 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=libwebm_libwebm-1.0.0.27_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=libwebm_libwebm-1.0.0.27_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [0 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=libwebm_libwebm-1.0.0.27_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=libwebm_libwebm-1.0.0.27_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| xerces | [1 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=xerces_v3.2.3_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=xerces_v3.2.3_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [3 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=xerces_v3.2.3_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=xerces_v3.2.3_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| bitcoin | [0 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=bitcoin_v0.20.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=bitcoin_v0.20.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [9 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=bitcoin_v0.20.1_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=bitcoin_v0.20.1_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| protobuf | [0 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=protobuf_v3.13.0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=protobuf_v3.13.0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [4 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=protobuf_v3.13.0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=protobuf_v3.13.0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
| qtbase | [16 new reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=qtbase_v6.2.0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=qtbase_v6.2.0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=New) | [49 resolved reports](https://codechecker-demo.eastus.cloudapp.azure.com/Default/reports?run=qtbase_v6.2.0_edonnag_Ericsson-assume-first-iteration_a4c7718&newcheck=qtbase_v6.2.0_edonnag_Ericsson-assume-first-iteration_b3c6379&diff-type=Resolved)
>From d938ccc8bf6a0498ea2bf0fad89f4c207e8813f0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Don=C3=A1t=20Nagy?= <donat.nagy at ericsson.com>
Date: Mon, 23 Dec 2024 12:20:35 +0100
Subject: [PATCH] [analyzer][NFC] Add option assume-one-iteration
This commit adds the new analyzer option `assume-one-iteration`, which
is `false` by default, but can be set to `true` to ensure that the
analyzer always assumes (at least) one iteration in loops.
In some situations this "loop is skipped" execution path is an
important corner case that may evade the notice of the developer and
hide significant bugs -- however, there are also many situations where
it's guaranteed that at least one iteration will happen (e.g. some
data structure is always nonempty), but the analyzer cannot realize
this and will produce false positives when it assumes that the loop is
skipped.
This commit refactors some logic around the implementation of the new
feature, but those changes are clearly NFC when the new analyzer option
is in its default state.
---
.../StaticAnalyzer/Core/AnalyzerOptions.def | 10 +++
clang/lib/StaticAnalyzer/Core/ExprEngine.cpp | 42 ++++++---
clang/test/Analysis/analyzer-config.c | 1 +
clang/test/Analysis/loop-assumptions.c | 88 ++++++++++++++-----
4 files changed, 106 insertions(+), 35 deletions(-)
diff --git a/clang/include/clang/StaticAnalyzer/Core/AnalyzerOptions.def b/clang/include/clang/StaticAnalyzer/Core/AnalyzerOptions.def
index 34bb7a809162ba..f98500f5fcf1dc 100644
--- a/clang/include/clang/StaticAnalyzer/Core/AnalyzerOptions.def
+++ b/clang/include/clang/StaticAnalyzer/Core/AnalyzerOptions.def
@@ -294,6 +294,16 @@ ANALYZER_OPTION(
bool, ShouldUnrollLoops, "unroll-loops",
"Whether the analysis should try to unroll loops with known bounds.", false)
+ANALYZER_OPTION(
+ bool, ShouldAssumeOneIteration, "assume-one-iteration",
+ "Whether the analyzer should always assume at least one iteration in "
+ "loops where the loop condition is opaque (i.e. the analyzer cannot "
+ "determine if it's true or false). Setting this to true eliminates some "
+ "false positives (where e.g. a structure is nonempty, but the analyzer "
+ "does not notice this); but it also eliminates some true positives (e.g. "
+ "cases where a structure can be empty and this causes buggy behavior).",
+ false)
+
ANALYZER_OPTION(
bool, ShouldDisplayNotesAsEvents, "notes-as-events",
"Whether the bug reporter should transparently treat extra note diagnostic "
diff --git a/clang/lib/StaticAnalyzer/Core/ExprEngine.cpp b/clang/lib/StaticAnalyzer/Core/ExprEngine.cpp
index 2b1872f8386aad..8c822a0ecc3b6f 100644
--- a/clang/lib/StaticAnalyzer/Core/ExprEngine.cpp
+++ b/clang/lib/StaticAnalyzer/Core/ExprEngine.cpp
@@ -2810,13 +2810,17 @@ void ExprEngine::processBranch(
if (StTrue && StFalse)
assert(!isa<ObjCForCollectionStmt>(Condition));
+ bool BothFeasible =
+ (StTrue && StFalse) ||
+ didEagerlyAssumeBifurcateAt(PrevState, dyn_cast<Expr>(Condition));
+
if (StTrue) {
- // If we are processing a loop condition where two iterations have
- // already been completed and the false branch is also feasible, then
- // don't assume a third iteration because it is a redundant execution
- // path (unlikely to be different from earlier loop exits) and can cause
- // false positives if e.g. the loop iterates over a two-element structure
- // with an opaque condition.
+ // In a loop, if both branches are feasible (i.e. the analyzer doesn't
+ // understand the loop condition) and two iterations have already been
+ // completed, then don't assume a third iteration because it is a
+ // redundant execution path (unlikely to be different from earlier loop
+ // exits) and can cause false positives if e.g. the loop iterates over a
+ // two-element structure with an opaque condition.
//
// The iteration count "2" is hardcoded because it's the natural limit:
// * the fact that the programmer wrote a loop (and not just an `if`)
@@ -2827,10 +2831,7 @@ void ExprEngine::processBranch(
// two iterations". (This pattern is common in FFMPEG and appears in
// many other projects as well.)
bool CompletedTwoIterations = IterationsCompletedInLoop.value_or(0) >= 2;
- bool FalseAlsoFeasible =
- StFalse ||
- didEagerlyAssumeBifurcateAt(PrevState, dyn_cast<Expr>(Condition));
- bool SkipTrueBranch = CompletedTwoIterations && FalseAlsoFeasible;
+ bool SkipTrueBranch = BothFeasible && CompletedTwoIterations;
// FIXME: This "don't assume third iteration" heuristic partially
// conflicts with the widen-loop analysis option (which is off by
@@ -2840,8 +2841,25 @@ void ExprEngine::processBranch(
Builder.generateNode(StTrue, true, PredN);
}
- if (StFalse)
- Builder.generateNode(StFalse, false, PredN);
+ if (StFalse) {
+ // In a loop, if both branches are feasible (i.e. the analyzer doesn't
+ // understand the loop condition), we are before the first iteration and
+ // the analyzer option `assume-one-iteration` is set to `true`, then avoid
+ // creating the execution path where the analyzer skips the loop.
+ //
+ // In some situations this "loop is skipped" execution path is an
+ // important corner case that may evade the notice of the developer and
+ // hide significant bugs -- however, there are also many situations where
+ // it's guaranteed that at least one iteration will happen (e.g. some
+ // data structure is always nonempty), but the analyzer cannot realize
+ // this and will produce false positives when it assumes that the loop is
+ // skipped.
+ bool BeforeFirstIteration = IterationsCompletedInLoop == std::optional{0};
+ bool SkipFalseBranch = BothFeasible && BeforeFirstIteration &&
+ AMgr.options.ShouldAssumeOneIteration;
+ if (!SkipFalseBranch)
+ Builder.generateNode(StFalse, false, PredN);
+ }
}
currBldrCtx = nullptr;
}
diff --git a/clang/test/Analysis/analyzer-config.c b/clang/test/Analysis/analyzer-config.c
index d5eb790b82f238..3611f6cb40c214 100644
--- a/clang/test/Analysis/analyzer-config.c
+++ b/clang/test/Analysis/analyzer-config.c
@@ -11,6 +11,7 @@
// CHECK-NEXT: alpha.osx.cocoa.DirectIvarAssignment:AnnotatedFunctions = false
// CHECK-NEXT: apply-fixits = false
// CHECK-NEXT: assume-controlled-environment = false
+// CHECK-NEXT: assume-one-iteration = false
// CHECK-NEXT: avoid-suppressing-null-argument-paths = false
// CHECK-NEXT: c++-allocator-inlining = true
// CHECK-NEXT: c++-container-inlining = false
diff --git a/clang/test/Analysis/loop-assumptions.c b/clang/test/Analysis/loop-assumptions.c
index eb0ffdce722e0c..bd8b74b1d531f9 100644
--- a/clang/test/Analysis/loop-assumptions.c
+++ b/clang/test/Analysis/loop-assumptions.c
@@ -1,25 +1,48 @@
// RUN: %clang_analyze_cc1 -analyzer-checker=debug.ExprInspection \
-// RUN: -verify=expected,eagerlyassume %s
+// RUN: -verify=expected,noassumeone,eagerlyassume,combo %s
// RUN: %clang_analyze_cc1 -analyzer-checker=debug.ExprInspection \
// RUN: -analyzer-config eagerly-assume=false \
+// RUN: -verify=expected,noassumeone,noeagerlyassume,combo %s
+// RUN: %clang_analyze_cc1 -analyzer-checker=debug.ExprInspection \
+// RUN: -analyzer-config assume-one-iteration=true \
+// RUN: -verify=expected,eagerlyassume,combo %s
+// RUN: %clang_analyze_cc1 -analyzer-checker=debug.ExprInspection \
+// RUN: -analyzer-config assume-one-iteration=true,eagerly-assume=false \
// RUN: -verify=expected,noeagerlyassume %s
+// The verify tag "combo" is used for one unique warning which is produced in three
+// of the four RUN combinations.
+
// These tests validate the logic within `ExprEngine::processBranch` which
// ensures that in loops with opaque conditions we don't assume execution paths
// if the code does not imply that they are possible.
+// In particular, if two (or more) iterations are already completed in a loop,
+// we don't assume that there can be another iteration. Moreover, if the
+// analyzer option `assume-one-iteration` is enabled, then we don't assume that
+// a loop can be skipped completely.
void clang_analyzer_numTimesReached(void);
-void clang_analyzer_warnIfReached(void);
void clang_analyzer_dump(int);
-void clearCondition(void) {
- // If the analyzer can definitely determine the value of the loop condition,
+void clearTrueCondition(void) {
+ // If the analyzer can definitely determine that the loop condition is true,
// then this corrective logic doesn't activate and the engine executes
// `-analyzer-max-loop` iterations (by default, 4).
- for (int i = 0; i < 10; i++)
+ int i;
+ for (i = 0; i < 10; i++)
clang_analyzer_numTimesReached(); // expected-warning {{4}}
- clang_analyzer_warnIfReached(); // unreachable
+ clang_analyzer_dump(i); // Unreachable, no reports.
+}
+
+void clearFalseCondition(void) {
+ // If the analyzer can definitely determine that the loop condition is false,
+ // then the loop is (obviously) skipped, even in `assume-one-iteration` mode.
+ int i;
+ for (i = 0; i > 10; i++)
+ clang_analyzer_numTimesReached(); // Unreachable, no report.
+
+ clang_analyzer_dump(i); // expected-warning {{0}}
}
void opaqueCondition(int arg) {
@@ -28,10 +51,13 @@ void opaqueCondition(int arg) {
// that more than two iterations are possible. (It _does_ imply that two
// iterations may be possible at least in some cases, because otherwise an
// `if` would've been enough.)
- for (int i = 0; i < arg; i++)
+ // Moreover, if `assume-one-iteration` is enabled, then assume at least one
+ // iteration.
+ int i;
+ for (i = 0; i < arg; i++)
clang_analyzer_numTimesReached(); // expected-warning {{2}}
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+ clang_analyzer_dump(i); // noassumeone-warning {{0}} expected-warning {{1}} expected-warning {{2}}
}
int check(void);
@@ -42,22 +68,26 @@ void opaqueConditionCall(int arg) {
// insert an assertion to guide the analyzer and rule out more than two
// iterations (so the analyzer needs to proactively avoid those unjustified
// branches).
- while (check())
+ int i = 0; // Helper to distinguish the the branches after the loop.
+ while (check()) {
clang_analyzer_numTimesReached(); // expected-warning {{2}}
+ i++;
+ }
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+ clang_analyzer_dump(i); // noassumeone-warning {{0}} expected-warning {{1}} expected-warning {{2}}
}
void opaqueConditionDoWhile(int arg) {
// Same situation as `opaqueCondition()` but with a `do {} while ()` loop.
// This is tested separately because this loop type is a special case in the
// iteration count calculation.
+ // Obviously, this loop guarantees that at least one iteration will happen.
int i = 0;
do {
clang_analyzer_numTimesReached(); // expected-warning {{2}}
} while (i++ < arg);
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+ clang_analyzer_dump(i); // expected-warning {{1}} expected-warning {{2}}
}
void dontRememberOldBifurcation(int arg) {
@@ -69,7 +99,7 @@ void dontRememberOldBifurcation(int arg) {
// by default), because the code remembered that there was a bifurcation on
// the first iteration of the loop and didn't realize that this is obsolete.
- // NOTE: The variable `i` is introduced to ensure that the iterations of the
+ // NOTE: The variable `i` is significant to ensure that the iterations of the
// loop change the state -- otherwise the analyzer stops iterating because it
// returns to the same `ExplodedNode`.
int i = 0;
@@ -78,10 +108,12 @@ void dontRememberOldBifurcation(int arg) {
i++;
}
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+ clang_analyzer_dump(i); // noassumeone-warning {{0}}
}
void dontAssumeFourthIterartion(int arg) {
+ int i;
+
if (arg == 2)
return;
@@ -89,10 +121,10 @@ void dontAssumeFourthIterartion(int arg) {
// iterations (because it knows that `arg != 2` at that point), so it
// performs a third iteration, but it does not assume that a fourth iteration
// is also possible.
- for (int i = 0; i < arg; i++)
+ for (i = 0; i < arg; i++)
clang_analyzer_numTimesReached(); // expected-warning {{3}}
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+ clang_analyzer_dump(i); // noassumeone-warning {{0}} expected-warning {{1}} expected-warning {{3}}
}
#define TRUE 1
@@ -108,20 +140,24 @@ void shortCircuitInLoopCondition(int arg) {
// false positive on the ffmpeg codebase. Eventually we should properly
// recognize the full syntactical loop condition expression as "the loop
// condition", but this will be complicated to implement.
- for (int i = 0; i < arg && TRUE; i++) {
+ int i;
+ for (i = 0; i < arg && TRUE; i++) {
clang_analyzer_numTimesReached(); // expected-warning {{4}}
}
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+
+ clang_analyzer_dump(i); // expected-warning {{0}} expected-warning {{1}} expected-warning {{2}} expected-warning {{3}}
}
void shortCircuitInLoopConditionRHS(int arg) {
// Unlike `shortCircuitInLoopCondition()`, this case is handled properly
// because the analyzer thinks that the right hand side of the `&&` is the
// loop condition.
- for (int i = 0; TRUE && i < arg; i++) {
+ int i;
+ for (i = 0; TRUE && i < arg; i++) {
clang_analyzer_numTimesReached(); // expected-warning {{2}}
}
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+
+ clang_analyzer_dump(i); // noassumeone-warning {{0}} expected-warning {{1}} expected-warning {{2}}
}
void eagerlyAssumeInSubexpression(int arg) {
@@ -134,16 +170,22 @@ void eagerlyAssumeInSubexpression(int arg) {
// sub-expression of the loop condition (as in this contrived test case).
// FIXME: I don't know a real-world example for this inconsistency, but it
// would be good to eliminate it eventually.
- for (int i = 0; (i >= arg) - 1; i++) {
+ int i;
+ for (i = 0; (i >= arg) - 1; i++) {
clang_analyzer_numTimesReached(); // eagerlyassume-warning {{4}} noeagerlyassume-warning {{2}}
}
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+
+ // The 'combo' warning intentionally appears when `assume-one-iteration` is
+ // disabled, but also appears as a bug (or at least inaccuracy) when
+ // `assume-one-iteration` is true but `EagerlyAssume` is also enabled.
+ clang_analyzer_dump(i); // combo-warning {{0}} expected-warning {{1}} expected-warning {{2}} eagerlyassume-warning {{3}}
}
void calledTwice(int arg, int isFirstCall) {
// This function is called twice (with two different unknown 'arg' values) to
// check the iteration count handling in this situation.
- for (int i = 0; i < arg; i++) {
+ int i;
+ for (i = 0; i < arg; i++) {
if (isFirstCall) {
clang_analyzer_numTimesReached(); // expected-warning {{2}}
} else {
@@ -215,5 +257,5 @@ void onlyLoopConditions(int arg) {
break;
}
- clang_analyzer_warnIfReached(); // expected-warning {{REACHABLE}}
+ clang_analyzer_dump(i); // expected-warning {{1}} expected-warning {{2}} expected-warning {{3}} expected-warning {{4}}
}
More information about the cfe-commits
mailing list