3rd argument in memcpy.

[mikeliaohm]

Hi, thank you for the great work.

I've encountered a seg fault after registering a new uri. The issue seems to related to memcpy of response_string. I think the 3rd argument should be the size of the response string to be copied instead of kMaxBufferSize, or whichever is smaller.

http-server/src/http_server.cc

Lines 238 to 242 in 1c17136

	response_string =
	to_string(http_response, http_request.method() != HttpMethod::HEAD);
	memcpy(raw_response->buffer, response_string.c_str(), kMaxBufferSize);
	raw_response->length = response_string.length();
	}

this is a potential fix.

  response_string
      = to_string (http_response, http_request.method () != HttpMethod::HEAD);
  const size_t res_len = response_string.length ();
  size_t len_to_copy = (res_len < kMaxBufferSize) 
    ? res_len : kMaxBufferSize;

  memmove (raw_response->buffer, response_string.c_str (), len_to_copy);
  
  raw_response->buffer[response_string.length ()] = '\0';
  raw_response->length = response_string.length ();

I've forked the project to reproduce my error at this repo.

[trungams]

Hi Mike, thank you for your interest in the repo, and thank you for your help identifying this bug. The previous code was using strncpy to copy the response content, but its behaviour is a bit different from memcpy, so we end up with the issue you described. I'll make a fix, or you can also submit a pull request if you want.

[mikeliaohm]

hi, thanks for the response. I've opened a PR. I also tried to properly null-terminate raw_response->buffer. See discussion in stackoverflow. Cheers.

[jmu201621143028]

Thanks for your discussion