[BOLT][BTI] Fix assertions checking getNumOperands

[bgergely0]

Several BTI-related functions are checking that a call MCInst has one
non-annotation operand.

This patch changes these checks to use MCPlus::getNumPrimeOperands,
instead of getNumOperands.

Testing: added annotations to existing gtests to serve as regression tests.
These now also explicitly check getNumOperands and getNumPrimeOperands
usage on the annotated MCInsts.

[llvmbot]

@llvm/pr-subscribers-bolt

Author: Gergely Bálint (bgergely0)

Changes

Use MCPlus::getNumPrimeOperands to check number of non-annotation
operands instead of getNumOperands.

---

Full diff: https://github.com/llvm/llvm-project/pull/174600.diff

1 Files Affected:

• (modified) bolt/lib/Target/AArch64/AArch64MCPlusBuilder.cpp (+2-2)

diff --git a/bolt/lib/Target/AArch64/AArch64MCPlusBuilder.cpp b/bolt/lib/Target/AArch64/AArch64MCPlusBuilder.cpp
index 03fb4ddc2f238..a30799d5f45d3 100644
--- a/bolt/lib/Target/AArch64/AArch64MCPlusBuilder.cpp
+++ b/bolt/lib/Target/AArch64/AArch64MCPlusBuilder.cpp
@@ -2818,7 +2818,7 @@ class AArch64MCPlusBuilder : public MCPlusBuilder {
     // x16 or x17. If the operand is not x16 or x17, it can be accepted by a BTI
     // j or BTI jc (and not BTI c).
     if (isIndirectBranch(Call)) {
-      assert(Call.getNumOperands() == 1 &&
+      assert(MCPlus::getNumPrimeOperands(Call) == 1 &&
              "Indirect branch needs to have 1 operand.");
       assert(Call.getOperand(0).isReg() &&
              "Indirect branch does not have a register operand.");
@@ -2856,7 +2856,7 @@ class AArch64MCPlusBuilder : public MCPlusBuilder {
     // x16 or x17. If the operand is not x16 or x17, it can be accepted by a
     // BTI j or BTI jc (and not BTI c).
     if (isIndirectBranch(Call)) {
-      assert(Call.getNumOperands() == 1 &&
+      assert(MCPlus::getNumPrimeOperands(Call) == 1 &&
              "Indirect branch needs to have 1 operand.");
       assert(Call.getOperand(0).isReg() &&
              "Indirect branch does not have a register operand.");

[peterwaller-arm]

Fix sounds reasonable though it makes me wonder if other uses of getNumOperands are wrong: what about the IsExplicitBTI definition, for example?

[bgergely0]

me wonder if other uses of getNumOperands are wrong: what about the IsExplicitBTI definition, for example?

that's a good point. Let me check other uses, and add a few relevant tests.

[bgergely0]

• [BOLT][BTI] Fix assertions checking getNumOperands #174600 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[paschalis-mpeis]

Looks good! Thanks Gergely, can you update the commit message to note that unit tests were updated to cover this?