KeyValueStoreMgr().Get() fails on Darwin

[rcasallas-silabs]

Problem

This behaviour has been observed on the Darwing implementation of KeyValueStoreMgr:

• If more than one new key is store with KeyValueStoreMgr().Put(), the following KeyValueStoreMgr().Get() call for the second key returns CHIP_NO_ERROR, but fails to retrieve the stored value (the output buffer remains unchanged).
• However, if KeyValueStoreMgr().Get() is called immediatelly after the KeyValueStoreMgr().Put() call, the value is later returned correctly.
• In any case, the KeyValueStoreMgr().Get() call does not modify the value of the in/out size parameter.
  The following code illustrate the problem:

static constexpr size_t kBufferSize = 512;
static constexpr size_t kEntriesCount = 4;
static constexpr size_t kEntrySizeMax = 128;
static constexpr size_t kKeySizeMax = 16;

void FormatKey(char *key, size_t max_size, size_t index)
{
    snprintf(key, max_size, "e%zu", index);
}

void TestKeyValueStoreMgr(nlTestSuite * apSuite, void * apContext)
{
    static uint8_t buffer[kBufferSize] = { 0 };
    ByteSpan entries[kEntriesCount];
    char key[kKeySizeMax];
    uint8_t out[kEntrySizeMax];
    size_t out_size;

    // Initialize

    for (size_t i = 0; i < sizeof(buffer); i++)
    {
        buffer[i] = static_cast<uint8_t>(i % 0xff);
    }
    
    // Delete

    for (size_t i = 0; i < kEntriesCount; i++)
    {
        FormatKey(key, sizeof(key), i);
        NL_TEST_ASSERT(apSuite, CHIP_NO_ERROR == chip::DeviceLayer::PersistedStorage::KeyValueStoreMgr().Delete(key));
    }

    // Save

    for (size_t i = 0; i < kEntriesCount; i++)
    {
        FormatKey(key, sizeof(key), i);
        entries[i] = ByteSpan(&buffer[16 + i], (1 + (i % kEntrySizeMax)));
        NL_TEST_ASSERT(apSuite, CHIP_NO_ERROR == chip::DeviceLayer::PersistedStorage::KeyValueStoreMgr().Put(key, entries[i].data(), entries[i].size()));
        out_size = sizeof(out);
        // FIXME: Loading will fail if the following line is commented
        // NL_TEST_ASSERT(apSuite, CHIP_NO_ERROR == chip::DeviceLayer::PersistedStorage::KeyValueStoreMgr().Get(key, out, out_size));
    }

    // Load

    for (size_t i = 0; i < kEntriesCount; i++)
    {
        FormatKey(key, sizeof(key), i);
        out_size = sizeof(out);
        memset(out, 0x00, out_size);
        NL_TEST_ASSERT(apSuite, CHIP_NO_ERROR == chip::DeviceLayer::PersistedStorage::KeyValueStoreMgr().Get(key, out, out_size));
        NL_TEST_ASSERT(apSuite, !memcmp(out, entries[i].data(), entries[i].size()));
    }
}

Proposed Solution

None so far.

[bzbarsky-apple]

@sagar-apple Is this code you know about? Or does @pan-apple ?

[bzbarsky-apple]

It's possible that #14233 will fix this. At least it fixes the above-pasted test for me....