[CIR] Handle the 'before case' block of a switch statement.#199752
Open
erichkeane wants to merge 2 commits into
Open
[CIR] Handle the 'before case' block of a switch statement.#199752erichkeane wants to merge 2 commits into
erichkeane wants to merge 2 commits into
Conversation
Before this patch, we would fail any time there was a block with entry/exit (in this case, one with successors thanks to a label) with a verification error. This patch adds special handling for that first block. This patch DOES choose to not trim them however. Unless there is a label inside of the block, there isn't any way to get there, and it is dead code. I've opted to NOT do that optimization, as I suspect that might be valuable to future passes/something we may wish to warn about in some sort of CFG analysis. Additionally, there is some minor changes to FlattenCFG, first to make sure we skip the switch ONLY if it is truly empty, and second to make sure we transform any 'break' in the pre-case region.
erichkeane
requested review from
andykaylor,
bcardosolopes,
lanza and
xlauko
as code owners
llvmorg-github-actions
Bot
added
clang
|
@llvm/pr-subscribers-clang @llvm/pr-subscribers-clangir Author: Erich Keane (erichkeane) ChangesBefore this patch, we would fail any time there was a block with entry/exit (in this case, one with successors thanks to a label) with a verification error. This patch adds special handling for that first block. This patch DOES choose to not trim them however. Unless there is a label inside of the block, there isn't any way to get there, and it is dead code. I've opted to NOT do that optimization, as I suspect that might be valuable to future passes/something we may wish to warn about in some sort of CFG analysis. Additionally, there is some minor changes to FlattenCFG, first to make sure we skip the switch ONLY if it is truly empty, and second to make sure we transform any 'break' in the pre-case region. Full diff: https://github.com/llvm/llvm-project/pull/199752.diff 3 Files Affected:
diff --git a/clang/lib/CIR/CodeGen/CIRGenStmt.cpp b/clang/lib/CIR/CodeGen/CIRGenStmt.cpp
index 4777d8e429e34..4a4380dea70b9 100644
--- a/clang/lib/CIR/CodeGen/CIRGenStmt.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenStmt.cpp
@@ -803,7 +803,7 @@ CIRGenFunction::emitCaseDefaultCascade(const T *stmt, mlir::Type condType,
// If the substmt is default stmt or case stmt, try to handle the special case
// to make it into the simple form. e.g.
//
- // swtich () {
+ // switch () {
// case 1:
// default:
// ...
@@ -1165,10 +1165,35 @@ mlir::LogicalResult CIRGenFunction::emitSwitchBody(const Stmt *s) {
auto *compoundStmt = cast<CompoundStmt>(s);
- mlir::Block *swtichBlock = builder.getBlock();
- for (auto *c : compoundStmt->body()) {
+ ArrayRef<Stmt *> body{compoundStmt->body_begin(), compoundStmt->body_end()};
+
+ mlir::Block *switchBlock = builder.getBlock();
+
+ // Any statements appearing before the first case statement are 'unassociated'
+ // with anything. So we have to create them FIRST in their own block. After
+ // that, the 'case' regions will take care of future ones.
+ if (!body.empty() && !isa<SwitchCase>(body.front())) {
+ builder.setInsertionPointToEnd(switchBlock);
+ while (!body.empty() && !isa<SwitchCase>(body.front())) {
+
+ auto *c = body.front();
+ if (mlir::failed(emitStmt(c, /*useCurrentScope=*/!isa<CompoundStmt>(c))))
+ return mlir::failure();
+
+ body = body.drop_front();
+ }
+
+ // Now that we've emitted ALL of the statements, we can create a new block
+ // for the actual case statements/etc to appear.
+ mlir::Block *lastBlock = builder.getBlock();
+ switchBlock = builder.createBlock(switchBlock->getParent());
+ builder.setInsertionPointToEnd(lastBlock);
+ cir::BrOp::create(builder, getLoc(s->getSourceRange()), switchBlock);
+ }
+
+ for (auto *c : body) {
if (auto *switchCase = dyn_cast<SwitchCase>(c)) {
- builder.setInsertionPointToEnd(swtichBlock);
+ builder.setInsertionPointToEnd(switchBlock);
// Reset insert point automatically, so that we can attach following
// random stmt to the region of previous built case op to try to make
// the being generated `cir.switch` to be in simple form.
diff --git a/clang/lib/CIR/Dialect/Transforms/FlattenCFG.cpp b/clang/lib/CIR/Dialect/Transforms/FlattenCFG.cpp
index a21394dc62332..8d77a18cb8c99 100644
--- a/clang/lib/CIR/Dialect/Transforms/FlattenCFG.cpp
+++ b/clang/lib/CIR/Dialect/Transforms/FlattenCFG.cpp
@@ -301,15 +301,16 @@ class CIRSwitchOpFlattening : public mlir::OpRewritePattern<cir::SwitchOp> {
if (hasNestedOpsToFlatten(region))
return mlir::failure();
- llvm::SmallVector<CaseOp> cases;
- op.collectCases(cases);
-
// Empty switch statement: just erase it.
- if (cases.empty()) {
+ if (op.getBody().hasOneBlock() &&
+ op.getBody().front().without_terminator().empty()) {
rewriter.eraseOp(op);
return mlir::success();
}
+ llvm::SmallVector<CaseOp> cases;
+ op.collectCases(cases);
+
// Create exit block from the next node of cir.switch op.
mlir::Block *exitBlock = rewriter.splitBlock(
rewriter.getBlock(), op->getNextNode()->getIterator());
@@ -322,6 +323,18 @@ class CIRSwitchOpFlattening : public mlir::OpRewritePattern<cir::SwitchOp> {
// 3. Replace the empty cir.switch.op with the new cir.switchflat op by the
// recorded block and conditions.
+ // First we have to handle the rewrite of all of the 'break' ops to make
+ // sure they now go to the right place, including the ones in the pre-case
+ // blcoks.
+ walkRegionSkipping<cir::LoopOpInterface, cir::SwitchOp>(
+ op.getBody(), [&](mlir::Operation *op) {
+ if (!isa<cir::BreakOp>(op))
+ return mlir::WalkResult::advance();
+
+ lowerTerminator(op, exitBlock, rewriter);
+ return mlir::WalkResult::skip();
+ });
+
// inline everything from switch body between the switch op and the exit
// block.
{
@@ -359,6 +372,8 @@ class CIRSwitchOpFlattening : public mlir::OpRewritePattern<cir::SwitchOp> {
mlir::Block *defaultDestination = exitBlock;
mlir::ValueRange defaultOperands = exitBlock->getArguments();
+
+
// Digest the case statements values and bodies.
for (cir::CaseOp caseOp : cases) {
mlir::Region ®ion = caseOp.getCaseRegion();
@@ -389,16 +404,6 @@ class CIRSwitchOpFlattening : public mlir::OpRewritePattern<cir::SwitchOp> {
break;
}
- // Handle break statements.
- walkRegionSkipping<cir::LoopOpInterface, cir::SwitchOp>(
- region, [&](mlir::Operation *op) {
- if (!isa<cir::BreakOp>(op))
- return mlir::WalkResult::advance();
-
- lowerTerminator(op, exitBlock, rewriter);
- return mlir::WalkResult::skip();
- });
-
// Track fallthrough in cases.
for (mlir::Block &blk : region.getBlocks()) {
if (blk.getNumSuccessors())
diff --git a/clang/test/CIR/CodeGen/switch-pre-case-stmts.cpp b/clang/test/CIR/CodeGen/switch-pre-case-stmts.cpp
new file mode 100644
index 0000000000000..0a9327417f807
--- /dev/null
+++ b/clang/test/CIR/CodeGen/switch-pre-case-stmts.cpp
@@ -0,0 +1,364 @@
+// RUN: %clang_cc1 -std=c++23 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --input-file=%t.cir %s --check-prefix=CIR
+// RUN: %clang_cc1 -std=c++23 -triple x86_64-unknown-linux-gnu -fclangir -emit-llvm %s -o %t-cir.ll
+// RUN: FileCheck --input-file=%t-cir.ll %s --check-prefix=LLVM
+// RUN: %clang_cc1 -std=c++23 -triple x86_64-unknown-linux-gnu -emit-llvm %s -o %t.ll
+// RUN: FileCheck --input-file=%t.ll %s --check-prefix=OGCG
+
+void pre_case_label() {
+ switch (1) {
+ LABEL1:;
+ case 1:;
+ }
+}
+
+// CIR-LABEL: cir.func{{.*}} @_Z14pre_case_labelv
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.br ^bb1
+// CIR: ^bb1:
+// CIR: cir.label "LABEL1"
+// CIR: cir.case(equal, [#cir.int<1> : !s32i]) {
+// CIR: cir.yield loc(#loc7)
+// CIR: }
+// CIR: cir.yield
+// CIR: }
+
+// LLVM-LABEL: define{{.*}} void @_Z14pre_case_labelv
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: i32 1, label %[[CASE1:.*]]
+// LLVM: ]
+// LLVM: [[LABEL:.*]]:{{.*}}; No predecessors!
+// LLVM: br label %[[JUMPS_TO_CASE1:.*]]
+// LLVM: [[JUMPS_TO_CASE1]]:
+// LLVM: br label %[[CASE1]]
+// CIR inserts a bunch of 'empty' blocks all over the place that just jump to
+// the next, so unifying these check blocks/checking more is pretty
+// awful/doesn't result in good checks.
+// LLVM: [[END]]:
+// LLVM: br label
+
+// OGCG-LABEL: define{{.*}} void @_Z14pre_case_labelv
+// OGCG: switch i32 1, label %[[END:.*]] [
+// OGCG: i32 1, label %[[CASE1:.*]]
+// OGCG: ]
+// OGCG: [[LABEL:.*]]:{{.*}}; No predecessors!
+// OGCG: br label %[[CASE1]]
+// OGCG: [[CASE1]]:
+// OGCG: br label %[[END]]
+// OGCG: [[END]]:
+// OGCG: ret void
+
+void multiple_pre_case_labels() {
+ switch (1) {
+ LABEL1:;
+ LABEL2:;
+ case 1:;
+ }
+}
+
+// CIR-LABEL: cir.func{{.*}} @_Z24multiple_pre_case_labelsv
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.br ^bb1
+// CIR: ^bb1:
+// CIR: cir.label "LABEL1"
+// CIR: cir.br ^bb2
+// CIR: ^bb2:
+// CIR: cir.label "LABEL2"
+// CIR: cir.case(equal, [#cir.int<1> : !s32i]) {
+// CIR: cir.yield
+// CIR: }
+// CIR: cir.yield
+// CIR: }
+
+// LLVM-LABEL: define{{.*}} @_Z24multiple_pre_case_labelsv
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: i32 1, label %[[CASE1:.*]]
+// LLVM: ]
+// LLVM: [[LABEL1:.*]]:
+// LLVM: br label %[[LABEL2:.*]]
+// LLVM: [[LABEL2]]:
+// LLVM: br label %[[JUMPS_TO_CASE1:.*]]
+// LLVM: [[JUMPS_TO_CASE1]]:
+// LLVM: br label %[[CASE1]]
+// LLVM: [[CASE1]]:
+// LLVM: br label %[[JUMPS_TO_END:.*]]
+// LLVM: [[JUMPS_TO_END]]:
+// LLVM: br label %[[END]]
+// LLVM: [[END]]:
+// LLVM: br label
+
+// OGCG-LABEL: define{{.*}} @_Z24multiple_pre_case_labelsv
+// OGCG: switch i32 1, label %[[END:.*]] [
+// OGCG: i32 1, label %[[CASE1:.*]]
+// OGCG: ]
+// OGCG: [[LABEL1:.*]]:{{.*}}; No predecessors!
+// OGCG: br label %[[LABEL2:.*]]
+// OGCG: [[LABEL2]]:
+// OGCG: br label %[[CASE1]]
+// OGCG: [[CASE1]]:
+// OGCG: br label %[[END]]
+// OGCG: [[END]]:
+// OGCG: ret void
+// OGCG: }
+
+void pre_case_goto() {
+ switch (1) {
+ goto end;
+ case 1:;
+ end:;
+ }
+}
+
+// CIR-LABEL: cir.func{{.*}} @_Z13pre_case_gotov
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.goto "end"
+// CIR: ^bb1:
+// CIR: cir.case(equal, [#cir.int<1> : !s32i]) {
+// CIR: cir.br ^bb1
+// CIR: ^bb1:
+// CIR: cir.label "end"
+// CIR: cir.yield
+// CIR: }
+// CIR: cir.yield
+// CIR: }
+
+// Classic-codegen manages to remove the switch, but we don't have that sort of
+// analysis working right. So this variant has a switch still in place.
+// LLVM-LABEL: define{{.*}} @_Z13pre_case_gotov
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: i32 1, label %[[CASE1:.*]]
+// LLVM: ]
+//
+// OGCG-LABEL: define{{.*}} @_Z13pre_case_gotov
+// OGCG: br label %[[END:.*]]
+// OGCG: [[END]]:
+// OGCG: ret void
+
+void pre_case_if(int cond) {
+ switch (1) {
+ if (cond) {}
+ case 1:;
+ }
+}
+
+// CIR-LABEL: cir.func{{.*}} @_Z11pre_case_ifi
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.scope {
+// CIR: cir.load
+// CIR: cir.cast int_to_bool
+// CIR: cir.if {{.*}} {
+// CIR: }
+// CIR: }
+// CIR: cir.case(equal, [#cir.int<1> : !s32i]) {
+// CIR: cir.yield
+// CIR: }
+// CIR: cir.yield
+// CIR: }
+
+// Once again, classic-codegen manages to make this a 'noop' and remove it.
+// LLVM-LABEL: define{{.*}} @_Z11pre_case_ifi
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: i32 1, label %[[CASE1:.*]]
+// LLVM: ]
+// LLVM: [[ENTRY:.*]]:{{.*}}; No predecessors!
+// LLVM: br label %[[IF_BLOCK:.*]]
+// LLVM: [[IF_BLOCK]]:
+// LLVM: load i32
+// LLVM: icmp ne i32
+// LLVM: br i1 %{{.*}}, label %[[TRUE:.*]], label %[[FALSE:.*]]
+// LLVM: [[TRUE]]:
+// LLVM: br label %[[FALSE]]
+// LLVM: [[FALSE]]:
+// LLVM: br label %[[JUMPS_TO_CASE1:.*]]
+// LLVM: [[JUMPS_TO_CASE1:.*]]:
+// LLVM: br label %[[CASE1]]
+// LLVM: [[CASE1]]:
+// LLVM: [[END]]:
+
+// OGCG-LABEL: define{{.*}} @_Z11pre_case_ifi
+// OGCG: alloca i32
+// OGCG: store i32 %{{.*}}, ptr %
+// OGCG: ret void
+
+void pre_case_return() {
+ switch (1) {
+ return;
+ case 1:;
+ }
+}
+
+// CIR-LABEL: cir.func{{.*}} @_Z15pre_case_returnv
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.return
+// CIR: ^bb1:
+// CIR: cir.case(equal, [#cir.int<1> : !s32i]) {
+// CIR: cir.yield
+// CIR: }
+// CIR: cir.yield
+// CIR: }
+
+// Once again, classic codegen skips this entirely, but CIR doesn't.
+// LLVM-LABEL: define{{.*}} @_Z15pre_case_returnv
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: i32 1, label %[[CASE1:.*]]
+// LLVM: ]
+// LLVM: [[ENTRY:.*]]:{{.*}}; No predecessors!
+// LLVM: ret void
+// LLVM: [[CASE1]]:
+// LLVM: [[END]]:
+
+// OGCG-LABEL: define{{.*}} @_Z15pre_case_returnv
+// OGCG: ret void
+
+void pre_case_break() {
+ switch (1) {
+ break;
+ case 1:;
+ }
+}
+
+// CIR-LABEL: cir.func{{.*}} @_Z14pre_case_breakv
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.break
+// CIR: ^bb1:
+// CIR: cir.case(equal, [#cir.int<1> : !s32i]) {
+// CIR: cir.yield
+// CIR: }
+// CIR: cir.yield
+// CIR: }
+
+// LLVM-LABEL: define{{.*}} @_Z14pre_case_breakv
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: i32 1, label %[[CASE1:.*]]
+// LLVM: ]
+// LLVM: [[ENTRY:.*]]:{{.*}}; No predecessors!
+// LLVM: br label %[[END]]
+// LLVM: [[CASE1]]:
+// LLVM: [[END]]:
+//
+// OGCG-LABEL: define{{.*}} @_Z14pre_case_breakv
+// OGCG: ret void
+
+void label_only_switch() {
+ switch (1) {
+ LABEL:;
+ }
+}
+
+// CIR-LABEL: cir.func{{.*}} @_Z17label_only_switchv
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.br ^bb1
+// CIR: ^bb1:
+// CIR: cir.label "LABEL"
+// CIR: cir.yield
+// CIR: }
+// CIR:}
+
+// LLVM-LABEL: define{{.*}} @_Z17label_only_switchv
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: ]
+// LLVM: [[LABEL:.*]]:{{.*}}; No predecessors!
+// LLVM: br label %[[JUMPS_TO_END:.*]]
+// LLVM: [[JUMPS_TO_END]]:
+// LLVM: br label %[[END]]
+
+// OGCG-LABEL: define{{.*}} @_Z17label_only_switchv
+// OGCG: switch i32 1, label %[[END:.*]] [
+// OGCG: ]
+// OGCG: [[LABEL:.*]]:{{.*}}; No predecessors!
+// OGCG: br label %[[END]]
+// OGCG: [[END]]:
+// OGCG: ret void
+
+void external_goto_into_pre_case(int cond) {
+ if (cond) goto LABEL;
+ switch (1) {
+ LABEL:;
+ case 1:;
+ }
+}
+// CIR-LABEL: cir.func{{.*}} @_Z27external_goto_into_pre_casei
+// CIR: cir.if {{.*}} {
+// CIR: cir.goto "LABEL"
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.br ^bb1
+// CIR: ^bb1:
+// CIR: cir.label "LABEL"
+// CIR: cir.case(equal, [#cir.int<1> : !s32i]) {
+// CIR: cir.yield
+// CIR: }
+// CIR: cir.yield
+// CIR: }
+
+// LLVM-LABEL: define{{.*}} @_Z27external_goto_into_pre_casei
+// LLVM: br i1 %{{.*}}, label %[[TRUE:.*]], label %[[FALSE:.*]]
+// LLVM: [[TRUE]]:
+// LLVM: br label %[[LABEL:.*]]
+// LLVM: [[FALSE]]:
+// Some empty blocks that just do jumps removed here.
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: i32 1, label %[[CASE1:.*]]
+// LLVM: ]
+// LLVM: [[ENTRY:.*]]:{{.*}}; No predecessors!
+// LLVM: br label %[[LABEL]]
+// LLVM: [[LABEL]]:
+// LLVM: br label %[[CASE1]]
+// LLVM: [[CASE1]]:
+// LLVM: [[END]]:
+
+// OGCG-LABEL: define{{.*}} @_Z27external_goto_into_pre_casei
+// OGCG: br i1 %{{.*}}, label %[[TRUE:.*]], label %[[FALSE:.*]]
+// OGCG: [[TRUE]]:
+// OGCG: br label %[[LABEL:.*]]
+// OGCG: [[FALSE]]:
+// OGCG: switch i32 1, label %[[END:.*]] [
+// OGCG: i32 1, label %[[CASE1:.*]]
+// OGCG: ]
+// OGCG: [[LABEL]]:
+// OGCG: br label %[[CASE1]]
+// OGCG: [[CASE1]]:
+// OGCG: br label %[[END]]
+// OGCG: [[END]]:
+// OGCG: ret void
+
+void external_goto_into_pre_case_empty(int cond) {
+ if (cond) goto LABEL;
+ switch (1) {
+ LABEL:;
+ }
+}
+// CIR-LABEL: cir.func{{.*}} @_Z33external_goto_into_pre_case_emptyi
+// CIR: cir.if {{.*}} {
+// CIR: cir.goto "LABEL"
+// CIR: cir.switch({{.*}} : !s32i) {
+// CIR: cir.br ^bb1
+// CIR: ^bb1:
+// CIR: cir.label "LABEL"
+// CIR: cir.yield
+// CIR: }
+
+// LLVM-LABEL: define{{.*}} @_Z33external_goto_into_pre_case_emptyi
+// LLVM: br i1 %{{.*}}, label %[[TRUE:.*]], label %[[FALSE:.*]]
+// LLVM: [[TRUE]]:
+// LLVM: br label %[[LABEL:.*]]
+// LLVM: [[FALSE]]:
+// Some empty blocks that just do jumps removed here.
+// LLVM: switch i32 1, label %[[END:.*]] [
+// LLVM: ]
+// LLVM: [[ENTRY:.*]]:{{.*}}; No predecessors!
+// LLVM: br label %[[LABEL]]
+// LLVM: [[LABEL]]:
+// LLVM: br label %[[END]]
+// LLVM: [[END]]:
+
+// OGCG-LABEL: define{{.*}} @_Z33external_goto_into_pre_case_emptyi
+// OGCG: br i1 %{{.*}}, label %[[TRUE:.*]], label %[[FALSE:.*]]
+// OGCG: [[TRUE]]:
+// OGCG: br label %[[LABEL:.*]]
+// OGCG: [[FALSE]]:
+// OGCG: switch i32 1, label %[[END:.*]] [
+// OGCG: ]
+// OGCG: [[LABEL]]:
+// OGCG: br label %[[END]]
+// OGCG: [[END]]:
+// OGCG: ret void
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
andykaylor
approved these changes
May 27, 2026
|
|
||
| mlir::Block *swtichBlock = builder.getBlock(); | ||
| for (auto *c : compoundStmt->body()) { | ||
| ArrayRef<Stmt *> body{compoundStmt->body_begin(), compoundStmt->body_end()}; |
Contributor
There was a problem hiding this comment.
Can you not use compoundStmt->body() here?
Contributor
Author
There was a problem hiding this comment.
No, thats an iterator_range. I could use that instead of array-ref for the range, but it doesn't have a 'drop-front' type function.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Before this patch, we would fail any time there was a block with entry/exit (in this case, one with successors thanks to a label) with a verification error. This patch adds special handling for that first block.
This patch DOES choose to not trim them however. Unless there is a label inside of the block, there isn't any way to get there, and it is dead code. I've opted to NOT do that optimization, as I suspect that might be valuable to future passes/something we may wish to warn about in some sort of CFG analysis.
Additionally, there is some minor changes to FlattenCFG, first to make sure we skip the switch ONLY if it is truly empty, and second to make sure we transform any 'break' in the pre-case region.