This change should be done for the zero-fill as well I believe in the do while part.

Thanks for working on this! Next step is to change/add testcases for you change, if you run ninja check-clang-cir you will find out what breaks. Because you are changing CIR code that we want to reflect in LLVM code, the testcase should check both IRs, see an example here: clang/test/CIR/CodeGen/abstract-cond.c

Is there any automation tool for clang-cir file check, I remember using an update_* script?

// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
// RUN: build/bin/FileCheck --check-prefix=CIR --input-file=%t.cir %s
// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-llvm -fno-clangir-call-conv-lowering %s -o %t.ll
// RUN: build/bin/FileCheck --check-prefix=LLVM --input-file=%t.ll %s

I have fixed the old broken test, for now adding the one from the original issue snippet:

void aggr_init() {
  int g = 5;
  int g_arr[5] = {1, 2, 3, g};
}

You might be able to get something using mlir/utils/generate-test-checks.py, but not perfect.

// RUN: build/bin/FileCheck --check-prefix=CIR --input-file=%t.cir %s

Please do like the status quo:

// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s

Done with the changes, let me know if anything else is needed.

Could you land this for me, the tests have passed?