[lldb] Ensure that TestMemoryCache.py reads allocated memory

[igorkudrin]

The test reads 400 bytes of memory above the local variable. If the stack is shallow, this can reach non-allocated space, resulting in a test failure.

[llvmbot]

@llvm/pr-subscribers-lldb

Author: Igor Kudrin (igorkudrin)

Changes

The test reads 400 bytes of memory above the local variable. If the stack is shallow, this can reach non-allocated space, resulting in a test failure.

---

Full diff: https://github.com/llvm/llvm-project/pull/151635.diff

1 Files Affected:

• (modified) lldb/test/API/functionalities/memory/cache/main.cpp (+7-1)

diff --git a/lldb/test/API/functionalities/memory/cache/main.cpp b/lldb/test/API/functionalities/memory/cache/main.cpp
index 44d85c5e9d2c3..a539e48c2cdf4 100644
--- a/lldb/test/API/functionalities/memory/cache/main.cpp
+++ b/lldb/test/API/functionalities/memory/cache/main.cpp
@@ -1,5 +1,11 @@
-int main ()
+int test()
 {
     int my_ints[] = {0x42};
     return 0; // Set break point at this line.
 }
+
+int main ()
+{
+    int dummy[100];
+    return test();
+}

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[DavidSpickett]

LGTM

I saw the same problem with another test case on Windows on Arm. In that case I was able to lower the amount of memory read, but here it needs to be some amount greater than the l2 cache line size which is 512:

  def MemCacheLineSize: Property<"memory-cache-line-size", "UInt64">,
    DefaultUnsignedValue<512>,
    Desc<"The memory cache line size">;

The test reads 201 * 4 bytes = 804. Starts from -100 from the array, so a 100 item buffer after it makes sense. There will be a few registers in each frame to cover the last 1 integer.

[igorkudrin]

Thanks!

There will be a few registers in each frame to cover the last 1 integer.

I counted my_ints itself for this last 1 integer. Yes, there are some registers stored in the past frames, but there is no point in guessing how many.