JIT: preserve local access type in relop assertion propagation

[EgorBo]

Copy the replacement local's node type along with its local and SSA numbers. This allows the resulting self-comparison to fold instead of retaining a truncated load of an int local.

Fixes #133859.

[azure-pipelines]

Azure Pipelines:
Successfully started running 5 pipeline(s).
11 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[Copilot]

🟢 Approval recommended

The fix and regression coverage address the reported JIT miscompilation.

Pull request overview

Fixes JIT assertion propagation by preserving the replacement local’s access type, preventing incorrect narrow loads and comparisons.

Changes:

• Preserves propagated local node types.
• Adds regression coverage for mismatched byte/int comparisons.

File summaries

File	Description
src/coreclr/jit/assertionprop.cpp	Preserves the replacement local access type during relop propagation.
src/tests/JIT/Regression_ro_2/Runtime_133859.cs	Adds regression tests for issue #133859.

Review details

• Files reviewed: 2/2 changed files
• Comments generated: 0
• Review effort level: Lite

[EgorBo]

PTAL @jakobbotsch, no diffs

[EgorBo]

Ping @jakobbotsch

[jakobbotsch]

What is the propagation that is happening here? Not totally sure that we would expect any propagation here in the first place given that it probably drops the normalize-on-store truncation that was happening implicitly?

[EgorBo]

@jakobbotsch I think it was actually correct, but I decided to just remove a bunch of code and inline optGlobalAssertionIsEqualOrNotEqual there. it also has some diffs because I removed unnecessary guards

    // Bail out if op1 is not side effect free. Note we'll be bashing it below, unlike op2.
    if ((op1->gtFlags & GTF_SIDE_EFFECT) != 0)
    {
        return nullptr;
    }

    if (!op1->OperIs(GT_LCL_VAR, GT_IND))
    {
        return nullptr;
    }

since we only work with VN there (global AP) and I deleted the path that relied on the o1 being side-effect free.

[jakobbotsch]

Nice cleanup!