[xunit-report] (version >= 3) stdout/sdterr duplicated when test case is marked with xfail

[lbrack]

This is a very minor issue but still. Basically when marking a test as xfail and generating an xunit report, the xunit's <system-out> and <system-err> tags are generated twice.

It is to be noted that:

• running the same test with the --runxfail do produce the correct output.
• I can reproduce the problem with all the 3.x versions but not with prior versions (<3)

pip list of the virtual environment you are using

$ pip list
Package        Version
-------------- -------
attrs          18.1.0
more-itertools 4.1.0
pip            10.0.1
pluggy         0.6.0
py             1.5.3
pytest         3.5.1
setuptools     39.1.0
six            1.11.0
wheel          0.31.1

pytest and operating system versions

• python 3.6.2 (but could also be reproduced on 2.7)
• pytest 3.5.1
• Darwin Kernel Version 17.5.0: Fri Apr 13 19:32:32 PDT 2018;RELEASE_X86_64 x86_64

Minimal example if possible

Running the following code with py.test <path> --junit-xml=xunit-report.xml

import sys
import pytest

@pytest.mark.xfail(reason="for kicks")
def test_fail():
    sys.stdout.write("XFAIL This is stdout\n")
    sys.stderr.write("XFAIL This is stderr\n")
    assert 0, "Could this be represented as an exemption"

generates the following xml (with duplicate outputs for stdout and stderr)

<?xml version="1.0" encoding="utf-8"?>
<testsuite errors="0" failures="0" name="pytest" skips="1" tests="1" time="0.072">
  <testcase classname="testspace.test_special_xunit" file="testspace/test_special_xunit.py" line="3" name="test_fail" time="0.0023894309997558594">
<skipped message="expected test failure">for kicks</skipped>
<system-out>XFAIL This is stdout</system-out>
<system-err>XFAIL This is stderr</system-err>
<system-out>XFAIL This is stdout</system-out>
<system-err>XFAIL This is stderr</system-err>
</testcase>
</testsuite>

[pytestbot]

GitMate.io thinks possibly related issues are #3100 (Cannot dynamically mark a test as xfail), #552 (Can't mark.xfail a callable test param), #2875 (Segmentation fault in versions >=3.1.1), #1546 (When an xfail marked test passes(xpass), the report.outcome value shows as 'failed'), and #2942 (pytest >= 3.2.4 raises an exception when pytest.mark.xfail() is used as a fixture param).

[nicoddemus]

Fixed in #3499