TRACE:DATA? times out in 0.3.0/master, works in 0.2

Hi there, I experience the following issue: The script below works in 0.2 but does not in 0.3.0 (and master d12ef43a576fa4a856f686796b75cceb0efe8e26):

import visa
rm = visa.ResourceManager('@py')
scp = rm.open_resource('TCPIP::10.54.231.16::INSTR')
scp.write("*RST")
scp.write("INIT:CONT OFF")
scp.write("INIT;*WAI")
scp.query("TRACE1:DATA? TRACE1")

In 0.2 the last query returns in under a second In 0.3.0 I get a timeout (or with indefinite timeout it just sits there)

I tried to git bisect the issue. It stopped working with an EOFError in 15b814171ea048653f5bcd4db625c4b855557ef3. This EOFError was adressed in the more recent commit 7d876c5e5c08334815a2648911316bd5c23fb57a. With that commit it times out:

Traceback (most recent call last):
  File "traceminimal.py", line 7, in <module>
    scp.query("TRACE1:DATA? TRACE1")
  File "/home/test/visatest/env3/lib/python3.5/site-packages/pyvisa/resources/messagebased.py", line 569, in query
    return self.read()
  File "/home/test/visatest/env3/lib/python3.5/site-packages/pyvisa/resources/messagebased.py", line 413, in read
    message = self._read_raw().decode(enco)
  File "/home/test/visatest/env3/lib/python3.5/site-packages/pyvisa/resources/messagebased.py", line 386, in _read_raw
    chunk, status = self.visalib.read(self.session, size)
  File "/home/test/visatest/pyvisa-py/pyvisa-py/highlevel.py", line 350, in read
    raise errors.VisaIOError(ret[1])
pyvisa.errors.VisaIOError: VI_ERROR_TMO (-1073807339): Timeout expired before operation completed.

[jake42]

It depends on chunk_size. Adding the following to the script above:

for chunk in range(256, 10240+256, 256):
    scp.chunk_size = chunk
    try:
        scp.query("TRACE1:DATA? TRACE1")
        print(chunk)
    except pyvisa.errors.VisaIOError:
        pass

Outputs:

256
512
768
1024
1280
1536
1792
2048
2304
2560
2816
3072
3328
3584
3840
4864
5376
5888
6144
6912
7936
8448
9216
9984

[ssouffri]

I also hit this bug in version 1.10.1 and can confirmed that the workaround works 👍. I discovered that the default chunk_size appears to be 20480 and wonder if this is a typo and was meant to be 2048?