[libc] Maintain proper alignment for the hermetic tests malloc

[jhuber6]

Summary

We use a bump pointer to implement malloc for the hermetic tests.
Currently, we bump the pointer up by any amount. This means that calling
malloc(1) will misalign the buffer so any following malloc(8)
accesses will not be aligned. This causes problems in architectures
which require alignment.

[sivachandra]

libc/test/UnitTest/HermeticTestUtils.cpp

68

Is it 8 that we want or is it alignof(uintptr_t)?

[jhuber6]

libc/test/UnitTest/HermeticTestUtils.cpp

68

Eight is just a good number. I think to be consistent with most implementations of malloc it's alignof(long double) or something. But I could change it to use that number since it's clearer to say "We want alignment for at least a 64-bit type"

[jhuber6]

Changing magic number

[sivachandra]

Using alignof(long double) is the best I think. Feel free to change to it.

libc/test/UnitTest/HermeticTestUtils.cpp

68

Nit: If you want to do long double, you should use a convenience var:

constexpr size_t ALIGNMENT = aligngof(long double);

[jhuber6]

In D149863#4319485, @sivachandra wrote:

Using alignof(long double) is the best I think. Feel free to change to it.

Was somewhat concerned using long double since it would double the wasted space for most tests. I wonder if it's possible to run asan or ubsan on the hermetic tests to catch stuff like this.

libc/test/UnitTest/HermeticTestUtils.cpp

 }
 void *memset(void *ptr, int value, size_t count) {
   return __llvm_libc::memset(ptr, value, count);
 }
 
 // This is needed if the test was compiled with '-fno-use-cxa-atexit'.
 int atexit(void (*func)(void)) { return __llvm_libc::atexit(func); }
 
+constexpr uint64_t ALIGNMENT = alignof(uintptr_t);
+
 void *malloc(size_t s) {
+  // Keep the bump pointer aligned on an eight byte boundary.
+  s = ((s + ALIGNMENT - 1) / ALIGNMENT) * ALIGNMENT;
   void *mem = ptr;
   ptr += s;
   return mem;
 }
 
 void free(void *) {}
 
 void *realloc(void *ptr, size_t s) {

libc/test/src/__support/CMakeLists.txt

 add_custom_target(libc-support-tests)
 
-# This test fails with a misaigned address on NVPTX.
-if(NOT LIBC_GPU_TARGET_ARCHITECTURE_IS_NVPTX)
-  add_libc_test(
-    blockstore_test
-    SUITE
-      libc-support-tests
-    SRCS
-      blockstore_test.cpp
-    DEPENDS
-      libc.src.__support.blockstore
-  )
-endif()
 
 add_libc_test(
   endian_test
   SUITE
     libc-support-tests
   SRCS
     endian_test.cpp
   DEPENDS