Don't pass nullptr into memcpy

[vitalybuka]

Summary

It's undefined according UBSAN.
Not sure which CL caused test failures, but seems writeBytes for empty buffer
should be OK.

[vitalybuka]

return early

[majnemer]

Did this get an LGTM? This change doesn't look right. The old code would return msf_error_code::insufficient_buffer if Data.size() was greater than zero.

[vitalybuka]

Not yet, it's for post-submitted review to fix UBSAN bots.
Looking.

[vitalybuka]

Majnemer, You are right
// here if Buffer is empty:
if (Offset > - Data.size())

return make_error<MSFError>(msf_error_code::insufficient_buffer);

so it's essentially if (Offset >0 || Data.size() >0 )

I should submit the first patch which I've changed in the last moment for "early return" :-)

[vitalybuka]

Ah, right, but I guess it's unintended one.
Then I am not sure if we want to change something here.

llvm/trunk/include/llvm/DebugInfo/MSF/ByteStream.h

public:
   Error readLongestContiguousChunk(uint32_t Offset,
                                    ArrayRef<uint8_t> &Buffer) const override {
     return ImmutableStream.readLongestContiguousChunk(Offset, Buffer);
   }
 
   uint32_t getLength() const override { return ImmutableStream.getLength(); }
 
   Error writeBytes(uint32_t Offset, ArrayRef<uint8_t> Buffer) const override {
+    if (Buffer.empty())
+      return Error::success();
+
     if (Data.size() < Buffer.size())
       return make_error<MSFError>(msf_error_code::insufficient_buffer);
     if (Offset > Buffer.size() - Data.size())
       return make_error<MSFError>(msf_error_code::insufficient_buffer);
 
     uint8_t *DataPtr = const_cast<uint8_t *>(Data.data());
     ::memcpy(DataPtr + Offset, Buffer.data(), Buffer.size());
     return Error::success();