[Mlir-commits] [mlir] [mlir] Make the split markers of splitAndProcessBuffer configurable. (PR #84765)

[Ingo Müller]

================
@@ -27,20 +29,22 @@ struct LogicalResult;
 using ChunkBufferHandler = function_ref<LogicalResult(
     std::unique_ptr<llvm::MemoryBuffer> chunkBuffer, raw_ostream &os)>;
 
-/// Splits the specified buffer on a marker (`// -----`), processes each chunk
-/// independently according to the normal `processChunkBuffer` logic, and writes
-/// all results to `os`.
+extern inline const char *const kDefaultSplitMarker = "// -----";
+
+/// Splits the specified buffer on a marker (`// -----` by default), processes
+/// each chunk independently according to the normal `processChunkBuffer` logic,
+/// and writes all results to `os`.
 ///
 /// This is used to allow a large number of small independent tests to be put
-/// into a single file. `enableSplitting` can be used to toggle if splitting
-/// should be enabled, e.g. to allow for merging split and non-split code paths.
-/// When `insertMarkerInOutput` is true, split markers (`//-----`) are placed
-/// between each of the processed output chunks.
+/// into a single file. The input split marker is configurable. If it is empty
+/// merging is disabled, which allows for merging split and non-split code
+/// paths. Output split markers (`//-----` by default) are placed between each
+/// of the processed output chunks.
 LogicalResult
 splitAndProcessBuffer(std::unique_ptr<llvm::MemoryBuffer> originalBuffer,
                       ChunkBufferHandler processChunkBuffer, raw_ostream &os,
-                      bool enableSplitting = true,
-                      bool insertMarkerInOutput = false);
+                      llvm::StringRef inputSplitMarker = kDefaultSplitMarker,
+                      llvm::StringRef outputSplitMarker = kDefaultSplitMarker);
----------------
ingomueller-net wrote:

No. Good catch! Fixed with latest commit.

https://github.com/llvm/llvm-project/pull/84765