Java: Use the shared BasicBlocks library. #19505

aschackmull · 2025-05-16T12:28:40Z

What the title says. Also, since BasicBlock is now no longer cast-compatible with ControlFlowNode there's no reason to use all the bb-prefixed names, so I've deprecated those and switched to the defaults provided by the shared library.

Copilot

Pull Request Overview

Migration to the shared BasicBlock library and deprecation of legacy bb* methods and predicates, updating call sites to use the new BasicBlock API.

Replaced legacy dominance and successor/predecessor methods (bbDominates, getABBSuccessor, etc.) with the shared-library methods (dominates, getASuccessor, etc.).
Updated imports, removed unused Dominance imports, and augmented qlpack.yml to depend on codeql/controlflow.
Introduced SuccessorType definitions, updated BasicBlocks.qll to use the shared BasicBlock implementation, and added change notes.

Reviewed Changes

Copilot reviewed 39 out of 39 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
java/ql/lib/semmle/code/java/security/PathSanitizer.qll	Swapped `bbDominates` for `dominates`.
java/ql/lib/semmle/code/java/security/AndroidWebViewCertificateValidationQuery.qll	Swapped `bbPostDominates` for `postDominates`.
java/ql/lib/semmle/code/java/dataflow/internal/rangeanalysis/SsaReadPositionSpecific.qll	Updated `idOf` to use `getFirstNode().getAstNode()`.
java/ql/lib/semmle/code/java/dataflow/internal/SsaImpl.qll	Removed `Dom` import; switched to `immediatelyDominates` and `getASuccessor`.
java/ql/lib/semmle/code/java/dataflow/internal/DataFlowUtil.qll	Replaced `getABBSuccessor` calls with `getASuccessor`.
java/ql/lib/semmle/code/java/dataflow/internal/DataFlowPrivate.qll	Swapped `bbIDominates` and `getABBSuccessor` for shared APIs.
java/ql/lib/semmle/code/java/dataflow/internal/BaseSSA.qll	Removed `Dom` import; switched to shared dominator/successor APIs.
java/ql/lib/semmle/code/java/dataflow/TypeFlow.qll	Replaced `getABBPredecessor` and `bbDominates` with new methods.
java/ql/lib/semmle/code/java/dataflow/RangeAnalysis.qll	Updated `idOf` and successor calls to use `getFirstNode`/`getASuccessor`.
java/ql/lib/semmle/code/java/dataflow/Nullness.qll	Replaced all `getABBSuccessor`/`getAST*` with new basic-block methods.
java/ql/lib/semmle/code/java/controlflow/internal/GuardsLogic.qll	Updated dominance checks to use `dominates` and new successor logic.
java/ql/lib/semmle/code/java/controlflow/UnreachableBlocks.qll	Replaced predecessor and dominance calls with shared methods.
java/ql/lib/semmle/code/java/controlflow/SuccessorType.qll	Added new `SuccessorType` hierarchy.
java/ql/lib/semmle/code/java/controlflow/Paths.qll	Swapped legacy bb* calls for shared successor/dominance methods.
java/ql/lib/semmle/code/java/controlflow/Guards.qll	Updated `getTestSuccessor`, dominance checks to new form.
java/ql/lib/semmle/code/java/controlflow/Dominance.qll	Deprecated old predicates, forwarding to new methods.
java/ql/lib/semmle/code/java/controlflow/BasicBlocks.qll	Imported and extended shared `BasicBlock` implementation.
java/ql/lib/qlpack.yml	Added `codeql/controlflow` dependency.
java/ql/lib/change-notes/2025-05-16-shared-basicblocks.md	Documented deprecation and migration to shared `BasicBlock`.
cpp/ql/lib/semmle/code/cpp/rangeanalysis/new/internal/semantic/SemanticCFG.qll	Renamed `bbDominates` to `dominates`.

Comments suppressed due to low confidence (3)

java/ql/lib/semmle/code/java/controlflow/BasicBlocks.qll:6

[nitpick] The Dominance import is now redundant since the shared BasicBlock API provides dominance predicates; consider removing this import to clean up unused dependencies.

import Dominance

java/ql/lib/semmle/code/java/controlflow/SuccessorType.qll:1

The new SuccessorType hierarchy introduced here doesn’t appear to have dedicated tests; consider adding unit tests to cover NormalSuccessor, ConditionalSuccessor, and ExceptionSuccessor handling.

import java

java/ql/lib/semmle/code/java/dataflow/internal/DataFlowPrivate.qll:86

The BasicBlock type isn’t aliased in this module; you should add private import java as J and class BasicBlock = J::BasicBlock; at the top to ensure the correct type is used.

BasicBlock getImmediateBasicBlockDominator(BasicBlock bb) {

java/ql/lib/semmle/code/java/controlflow/BasicBlocks.qll

java/ql/lib/semmle/code/java/controlflow/Dominance.qll

 }

 /** Exit points for basic-block control-flow. */
 private predicate bbSink(BasicBlock exit) { exit.getLastNode() instanceof ControlFlow::ExitNode }

 /** Reversed `bbSucc`. */
-private predicate bbPred(BasicBlock post, BasicBlock pre) { post = pre.getABBSuccessor() }
+private predicate bbPred(BasicBlock post, BasicBlock pre) { post = pre.getASuccessor() }


paldepind

Really awesome to get Java onto to the shared basic block library 😃

I've left a few minor comments and ideas.

paldepind · 2025-05-20T08:19:40Z

java/ql/lib/semmle/code/java/controlflow/BasicBlocks.qll

-  /** Holds if this basic block post-dominates `node`. (This is reflexive.) */
-  predicate bbPostDominates(BasicBlock node) { bbPostDominates(this, node) }
+/**
+ * A control-flow node that represents the start of a basic block.


This comment seems outdated as BasicBlock is no longer also a ControlFlowNode.

What about using the exact same QLDoc as for BasicBlock in the other languages (which we could modify if needed)?

paldepind · 2025-05-20T08:23:14Z

java/ql/lib/change-notes/2025-05-16-shared-basicblocks.md

+---
+category: deprecated
+---
+* Java now uses the shared `BasicBlock` library. This means that several member predicates now use the preferred names. The old predicates have been deprecated. The `BasicBlock` class itself no longer extends `ControlFlowNode` - the predicate `getFirstNode` can be used to fix any QL code that somehow relied on this.


I'm not sure if it's clear for end users what "the preferred names" refer to? Maybe mention that the naming of the member predicates are now aligned with other languages, which also gives it a positive spin.

paldepind · 2025-05-20T08:27:13Z

java/ql/lib/semmle/code/java/controlflow/Dominance.qll

 * ```
 */
 predicate dominanceFrontier(BasicBlock x, BasicBlock w) {
-  x = w.getABBPredecessor() and not bbIDominates(x, w)
+  x = w.getAPredecessor() and not x.immediatelyDominates(w)


What about deprecating this in favor of inDominanceFrontier from the shared library?

I don't think dominance frontier is used for anything else than SSA construction, and the shared SSA lib does its own calculation, so perhaps we should deprecate both. In any case, I think that's fine to postpone for now.

paldepind · 2025-05-20T09:07:47Z

java/ql/lib/semmle/code/java/controlflow/Dominance.qll


 /** The immediate post-dominance relation on basic blocks. */
-cached
-predicate bbIPostDominates(BasicBlock dominator, BasicBlock node) =
+deprecated predicate bbIPostDominates(BasicBlock dominator, BasicBlock node) =


Should this have a DEPRECATED: Use ... doc like the other deprecated predicates?

Immediate post-dominance is not exposed in the shared lib - I'm not sure we've ever seen any use-cases for it, though, so adding it is probably of little value - hence why I simply opted to discretely deprecate this single public instance.

paldepind · 2025-05-20T11:27:12Z

shared/controlflow/codeql/controlflow/BasicBlock.qll

+   * This is a necessary and sufficient condition for an edge to dominate anything,
+   * and in particular `dominatingEdge(bb1, bb2) and bb2.dominates(bb3)` means
+   * that the edge `(bb1, bb2)` dominates `bb3`.


Suggested change

* This is a necessary and sufficient condition for an edge to dominate anything,

* and in particular `dominatingEdge(bb1, bb2) and bb2.dominates(bb3)` means

* that the edge `(bb1, bb2)` dominates `bb3`.

* This is a necessary and sufficient condition for an edge to dominate some

* node, and `dominatingEdge(bb1, bb2) and bb2.dominates(bb3)` implies that

* the edge `(bb1, bb2)` dominates `bb3`.

It might just be me, but I find "anything" here to be a bit unclear.

I've adopted the s/anything/some node/, but I don't want to change the second sentence from "means" to "implies", as that weakens the statement - it really is the semantic meaning of the QL snippet that's being explained.

Or how about "something" instead? Or at the very least "some block"?

I've gone with "some block" for now.

"Some block" is also great 👍

I don't really understand the "in particular" though. What follows doesn't seem like a special case of the preceding statement? "This" in the first part refers to dominatingEdge(bb1, bb2) and the second part talks about a situration where, in addition, bb2.dominates(bb3) holds as well?

Right, I guess "in particular" ought to be "therefore". Does that sound better to you?

Yes, I think "therefore" would be great :)

paldepind · 2025-05-20T11:33:43Z

java/ql/test/library-tests/controlflow/dominance/dominanceWrong.ql

@@ -16,6 +16,6 @@ predicate dominanceCounterExample(ControlFlowNode entry, ControlFlowNode dom, Co

 from Callable c, ControlFlowNode dom, ControlFlowNode node
 where
-  (strictlyDominates(dom, node) or bbStrictlyDominates(dom, node)) and
+  strictlyDominates(dom, node) and


This is in a file that includes "wrong" in the name, so I'm guessing it doesn't matter, but it's not clear to me why the previous code had the two conjuncts.

Once upon a time dominance for individual cfg nodes was computed independently of basic blocks, which I believe is the historical reason for this double check. Now, when translating basic block dominance via getFirstNode, one implies the other trivially, so we can simplify the test.

paldepind · 2025-05-20T11:36:16Z

java/ql/lib/semmle/code/java/controlflow/BasicBlocks.qll

+ */
+class BasicBlock extends BbImpl::BasicBlock {
+  /** Gets the immediately enclosing callable whose body contains this node. */
+  Callable getEnclosingCallable() { result = this.getScope() }


As discussed offline it might be nice to agree on a name for this predicate that every language can use. Right now Java's basic block will have both getScope and getEnclosingCallable that does the exact same thing.

Yes, but I think that's best left as a follow-up.

Yes, absolutely :)

aschackmull · 2025-05-20T12:31:53Z

@paldepind I've pushed review fixes now.

…cessor relation.

paldepind

Thanks for addressing my nitpicks. Looks great to me 👍

Copilot AI review requested due to automatic review settings May 16, 2025 12:28

aschackmull requested review from a team as code owners May 16, 2025 12:28

github-actions bot added C++ documentation Java Kotlin labels May 16, 2025

Copilot AI reviewed May 16, 2025

View reviewed changes

github-advanced-security bot found potential problems May 16, 2025

View reviewed changes

aschackmull force-pushed the java/basicblock branch 2 times, most recently from 5c50666 to 719bbe4 Compare May 19, 2025 07:37

paldepind requested changes May 20, 2025

View reviewed changes

aschackmull force-pushed the java/basicblock branch from 0514624 to 00073b3 Compare May 20, 2025 19:56

aschackmull added 7 commits May 21, 2025 09:01

Java: Use the shared BasicBlocks library.

f202586

Shared: Refactor the shared BasicBlock lib slightly and cache the suc…

13c5906

…cessor relation.

Java: Deprecate redundant basic block predicates.

db01828

Java: Add change note.

6b830fa

Java: Override dominates to reference the right type.

a98d93b

Java: Extend qldoc.

3fde675

Java/Shared: Address review comments.

10efea1

aschackmull force-pushed the java/basicblock branch from 00073b3 to 10efea1 Compare May 21, 2025 07:02

paldepind approved these changes May 21, 2025

View reviewed changes

aschackmull merged commit 00c7bc1 into github:main May 21, 2025
43 checks passed

aschackmull deleted the java/basicblock branch May 21, 2025 11:37

Java: Use the shared BasicBlocks library. #19505

Java: Use the shared BasicBlocks library. #19505

Uh oh!

Conversation

aschackmull commented May 16, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Uh oh!

Uh oh!

Check warning

paldepind left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

aschackmull commented May 20, 2025

Uh oh!

paldepind left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.