zopefoundation / zope.testbrowser Goto Github PK

View Code? Open in Web Editor NEW

21.0 71.0 16.0 1.67 MB

Programmable browser for functional black-box tests

License: Other

Python 93.70% HTML 6.30%

testing browser python maintained hacktoberfest

zope.testbrowser's People

Contributors

Stargazers

Watchers

Forkers

teixas nidico andrewsmedina flavianmissi jean fschulze kaymccormick nazrulworld cjwatson angry-boss python-repository-hub william3johnson inferno-inc

zope.testbrowser's Issues

Incomplete interface definitons

IControl doesn't define readonly, which looks like part of the API. (Although note that IListControl classes don't have readonly properties.)

IControl.type has a list of values that doesn't mention textarea. I also wonder how it would cope with HTML5 control types such as date etc.

IControl doesn't have id (and the implementation lacks it as well), which is sometimes rather inconvenient.

Test suite fails (cannot import name pystone)

Building zope.testbrowser on my Fedora 18 machine with python-2.7.3 and virtualenv-1.9.1 from PyPI fails, using the enclosed buildout. This came up in #4 originally. This is the full stack trace:

nico@pesto $ bin/test
Running zope.testrunner.layer.UnitTests tests:
  Set up zope.testrunner.layer.UnitTests in 0.000 seconds.


Failure in test /home/nico/src/zope.testbrowser/src/zope/testbrowser/README.txt
Failed doctest test for README.txt
  File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/README.txt", line 0

----------------------------------------------------------------------
File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/README.txt", line 119, in README.txt
Failed example:
    verifyObject(interfaces.IBrowser, browser)
Exception raised:
    Traceback (most recent call last):
      File "/usr/lib64/python2.7/doctest.py", line 1289, in __run
        compileflags, 1) in test.globs
      File "<doctest README.txt[20]>", line 1, in <module>
        verifyObject(interfaces.IBrowser, browser)
      File "/home/nico/src/zope.testbrowser/eggs/zope.interface-4.0.5-py2.7-linux-x86_64.egg/zope/interface/verify.py", line 105, in verifyObject
        return _verify(iface, candidate, tentative, vtype='o')
      File "/home/nico/src/zope.testbrowser/eggs/zope.interface-4.0.5-py2.7-linux-x86_64.egg/zope/interface/verify.py", line 55, in _verify
        attr = getattr(candidate, name)
      File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/browser.py", line 299, in lastRequestPystones
        return self.timer.elapsedPystones
      File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/browser.py", line 176, in elapsedPystones
        return self.elapsedSeconds * self.pystonesPerSecond
      File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/browser.py", line 135, in pystonesPerSecond
        from test import pystone
    ImportError: cannot import name pystone
----------------------------------------------------------------------
File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/README.txt", line 1380, in README.txt
Failed example:
    browser.lastRequestPystones < 10000 # really big number for safety
Exception raised:
    Traceback (most recent call last):
      File "/usr/lib64/python2.7/doctest.py", line 1289, in __run
        compileflags, 1) in test.globs
      File "<doctest README.txt[388]>", line 1, in <module>
        browser.lastRequestPystones < 10000 # really big number for safety
      File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/browser.py", line 299, in lastRequestPystones
        return self.timer.elapsedPystones
      File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/browser.py", line 176, in elapsedPystones
        return self.elapsedSeconds * self.pystonesPerSecond
      File "/home/nico/src/zope.testbrowser/src/zope/testbrowser/browser.py", line 135, in pystonesPerSecond
        from test import pystone
    ImportError: cannot import name pystone

  Ran 19 tests with 1 failures and 0 errors in 0.375 seconds.
Tearing down left over layers:
  Tear down zope.testrunner.layer.UnitTests in 0.000 seconds.

.reload() does not follow redirects nor does it raise exceptions for bad HTTP statuses

When you load a page and then reload it and the server sends a 30x redirect for the 2nd response, the .reload() call will not follow redirects (nor raise exceptions for 4xx/5xx errors).

This is a regression from mechanize-based zope.testbrowser.

Careless use of CSS selectors breaks with new BeautifulSoup

zope.testbrowser relies on BeautifulSoup's implementation of CSS selectors to query HTML nodes. The latest release of BeautifulSoup requires CSS selectors to follow the CSS spec in ways where previously it allowed lax values. Since zope.testbrowser doesn't actually follow the CSS spec (inadequate quoting), this causes problems like zopefoundation/zc.catalog#8.

One instance (from the above bug) is this line in getControlLabels():

        forlbls = html.select('label[for=%s]' % controlid)

The %s should apparently be quoted with double quotes in case controlid is not a valid CSS identifier.

Other instances of this may lurk. https://facelessuser.github.io/soupsieve/#beautiful-soup-differences has a list of things we should watch for.

installation via pip broken

To reproduce:

$ pip install zope.testbrowser==4.0.2

Output is:

Downloading/unpacking zope.testbrowser==4.0.2
  Downloading zope.testbrowser-4.0.2.tar.gz (77kB): 77kB downloaded
  Running setup.py egg_info for package zope.testbrowser

Downloading/unpacking mechanize>=0.2.0 (from zope.testbrowser==4.0.2)
  Downloading mechanize-0.2.5.tar.gz (383kB): 383kB downloaded
  Running setup.py egg_info for package mechanize

Requirement already satisfied (use --upgrade to upgrade): setuptools in /Users/flavia.carlette/.virtualenvs/test/lib/python2.7/site-packages/setuptools-0.6c11-py2.7.egg (from zope.testbrowser==4.0.2)
Downloading/unpacking zope.interface (from zope.testbrowser==4.0.2)
  Downloading zope.interface-4.0.5.zip (173kB): 173kB downloaded
  Running setup.py egg_info for package zope.interface

    warning: no previously-included files matching '*.dll' found anywhere in distribution
    warning: no previously-included files matching '*.pyc' found anywhere in distribution
    warning: no previously-included files matching '*.pyo' found anywhere in distribution
    warning: no previously-included files matching '*.so' found anywhere in distribution
Downloading/unpacking zope.schema (from zope.testbrowser==4.0.2)
  Downloading zope.schema-4.3.2.zip (93kB): 93kB downloaded
  Running setup.py egg_info for package zope.schema

    warning: no previously-included files matching '*.dll' found anywhere in distribution
    warning: no previously-included files matching '*.pyc' found anywhere in distribution
    warning: no previously-included files matching '*.pyo' found anywhere in distribution
    warning: no previously-included files matching '*.so' found anywhere in distribution
Downloading/unpacking pytz (from zope.testbrowser==4.0.2)
  Could not find a version that satisfies the requirement pytz (from zope.testbrowser==4.0.2) (from versions: 2009r, 2008b, 2009f, 2008c, 2007g, 2011g, 2005m, 2011e, 2007f, 2011k, 2007k, 2006j, 2008h, 2008i, 2011e, 2008a, 2009e, 2006g, 2011j, 2010l, 2005m, 2008i, 2005k, 2008g, 2007c, 2007i, 2009l, 2009r, 2006j, 2011k, 2007d, 2006p, 2009i, 2009u, 2007i, 2009f, 2010g, 2008h, 2009a, 2007g, 2011e, 2006p, 2012b, 2010k, 2005r, 2007f, 2009l, 2009p, 2008c, 2009j, 2008g, 2010g, 2010h, 2011h, 2010k, 2007c, 2007d, 2011d, 2009l, 2011c, 2008a, 2005m, 2007k, 2009n, 2011d, 2013b, 2012h, 2010e, 2012d, 2012f, 2011n, 2011b, 2011j, 2008c, 2012j, 2007k, 2009f, 2009d, 2010e, 2010b, 2013b, 2011d, 2009p, 2008h, 2005r, 2009i, 2009n, 2009a, 2010k, 2008g, 2006g, 2008b, 2012c, 2009i, 2007g, 2012c, 2010h, 2011n, 2012g, 2007d, 2008a, 2009u, 2012g, 2010o, 2006p, 2010b, 2009u, 2012d, 2011k, 2012f, 2009a, 2007f, 2011h, 2010l, 2009j, 2011g, 2009g, 2009g, 2005r, 2011c, 2012g, 2009g, 2012d, 2009j, 2010o, 2007c, 2010g, 2006g, 2009d, 2010h, 2005k, 2006j, 2010b, 2009n, 2011g, 2011c, 2012b, 2009e, 2009d, 2011j, 2007i, 2012j, 2010l, 2009r, 2012h, 2010e, 2009p, 2008i, 2012f, 2009e, 2012b, 2011h, 2005k, 2008b, 2012c, 2010o, 2013b, 2012j, 2011n, 2004b)
Cleaning up...
No distributions matching the version for pytz (from zope.testbrowser==4.0.2)

Installing pytz==2013b and then trying to install zope.testbrowser again fixes the issue.

Cannot use getLink ob Labels with shy-markers (soft hyphen)

The soft hyphen or -marker tell a web browser where it can break the word to start a new line and inserts a hyphen at the end of the splitted word. For example, article may be rendered as article or ar-\nticle or artic-\nle.

Since zope.testbrowser renders special markers like  or  , it would be necessary to insert a soft-hyphen into the expected text, so getLink finds the text. However, how do you represent a special marker that renders to an empty string?

zope.testbrowser should probably just remove  markers when searching for a link or control.

Latest stable release (4.0.3) incompatible with latest stable WebTest (any version > 1.3.4)

The WebTest <= 1.3.4 version pin in latest stable zope.testbrowser release (4.0.3) is causing pain.

I thought pushing zope.testbrowser 5.0.0 out of the door would be a good way to fix this, but on second though, that would cause even more pain due to lack of 100% backwards compatibility.

So let's fix this and release 4.0.4.

The version pin exists because zope.testbrowser's tests fail when you get a newer WebTest. The test failures are all of the form

AssertionError: Iterator <listiterator object at 0x37d5e10> returned a non-<type 'str'> object: u"Hello world!\n\nHTTP_CONNECTION = 'close'\nHTTP...

in webtest.lint, and are caused by stdlib's wsgiref.simple_server.demo_app returning a Unicode response body. Why does it return a unicode body? Because it prints the wsgi environment into a StringIO. Why are there Unicode strings in the wsgi environment? Frog knows, but here they are:

(Pdb) pp environ
{'HTTP_CONNECTION': 'close',
 'HTTP_HOST': 'localhost',
 'HTTP_USER_AGENT': 'Python-urllib/2.7',
 'PATH_INFO': '/',
 'QUERY_STRING': '',
 'REQUEST_METHOD': 'GET',
 'SCRIPT_NAME': '',
 'SERVER_NAME': 'localhost',
 'SERVER_PORT': '80',
 'SERVER_PROTOCOL': 'HTTP/1.0',
 u'paste.testing': True,
 u'paste.testing_variables': {},
 u'paste.throw_errors': None,
 'wsgi.errors': <webtest.lint.ErrorWrapper object at 0x7fdf4f6bbe10>,
 'wsgi.input': <webtest.lint.InputWrapper object at 0x7fdf4f628090>,
 'wsgi.multiprocess': False,
 'wsgi.multithread': False,
 'wsgi.run_once': False,
 'wsgi.url_scheme': 'http',
 'wsgi.version': (1, 0)}

getLink now tests regexes using .match, not .search

(I see several open issues about compatibility with the old mechanize-based test browser, so I hope this sort of thing is welcome.)

I'm in the process of migrating a large project (Launchpad) from old versions of zope.testbrowser to version 5.x. One thorn in my side has been that in several places we do something like this:

    edit_status_url = re.compile(r'1.0/\+bug/[0-9]+/\+editstatus')
    browser.getLink(url=edit_status_url).click()

This worked with mechanize because regexes were tested using .search. However, zope.testbrowser now tests them using .match, so the above code has to be more like this:

    edit_status_url = re.compile(r'.*/1.0/\+bug/[0-9]+/\+editstatus')
    browser.getLink(url=edit_status_url).click()

I don't suppose it would be possible to go back to 4.x-compatible behaviour here?

Python 3.8 support

Python 3.8 removed time.clock(), which causes

Traceback (most recent call last):
...
  Module zope.testbrowser.browser, line 483, in _preparedRequest
    self.timer.start()
  Module zope.testbrowser.browser, line 1420, in start
    self.start_time = self._getTime()
  Module zope.testbrowser.browser, line 1413, in _getTime
    return time.clock()
AttributeError: module 'time' has no attribute 'clock'

Value of unchecked checkboxes changed from False to [] in zope.testbrowser 5.0.0

The old mechanize-based zope.testbrowser behaved like this:

>>> print(browser.contents)
 ...
<input type="checkbox" id="closed" name="closed" value="on" />
...
>>> browser.getControl(name='closed').value
False

The new webtest-based zope.testbrowser returns [] instead of False. Which is fine, I suppose, but I'd like this backwards incompatibility to be documented somewhere.

new webtest broke our tests

See #29 (comment)

getControl('Label') cannot find hidden inputs any more

Another of those mechanize-to-webtest compatibility issues:

>>> print(browser.contents)
 ...<label for="form.lead_reviewer">Reviewer</label>
<input class="hiddenType" id="form.lead_reviewer" name="form.lead_reviewer" type="hidden" value=""  />
...
>>> browser.getControl('Reviewer').value = 'manager'

This used to work, now it fails with a LookupError.

(I'm fine with the change, <label> pointing to a hidden field seems wrong, I'd just like this to be documented.)

.reload() fails if the previous request was a POST

I've seen variations on this, some of which end up in an error like

Failed example:
    manager.reload()
Exception raised:
    Traceback (most recent call last):
      File "/usr/lib/python2.7/doctest.py", line 1315, in __run
        compileflags, 1) in test.globs
      File "<doctest invitations.txt[101]>", line 1, in <module>
        manager.reload()
      File "/home/mg/src/zopefoundation/zope.testbrowser/src/zope/testbrowser/browser.py", line 197, in reload
        resp = self.testapp.request(req)
      File "/home/mg/src/ivija/.tox/py27-latest/local/lib/python2.7/site-packages/webtest/app.py", line 581, in request
        req = url_or_req.copy()
      File "/home/mg/src/ivija/.tox/py27-latest/local/lib/python2.7/site-packages/webob/request.py", line 861, in copy
        self.make_body_seekable()
      File "/home/mg/src/ivija/.tox/py27-latest/local/lib/python2.7/site-packages/webob/request.py", line 929, in make_body_seekable
        self.copy_body()
      File "/home/mg/src/ivija/.tox/py27-latest/local/lib/python2.7/site-packages/webob/request.py", line 953, in copy_body
        data = input.read(min(todo, 65535))
      File "/home/mg/src/ivija/.tox/py27-latest/local/lib/python2.7/site-packages/webob/request.py", line 1544, in readinto
        "(%d more bytes were expected)" % (self.remaining,)
    DisconnectionError: The client disconnected while sending the body (79 more bytes were expected)

~~and some of which silently end up with a browser.contents of ''.~~ (that was actually a different bug, see #75)

I get the impression that submitting a form and then calling reload() in the old mechanize-based testbrowser used to replay the last POST request as a GET to the same URL.

I don't think that's a good idea either, so what I propose instead is to have zope.testbrowser explicitly raise a BrowserStateError() if you try to reload a POST.

Bug since switch from mechanize

The behaviour related to setting radio controls has changed in some cases.

I added a test case in master...fschulze:radio-master.

I added the same one for 4.0.4 in 4.0.4...fschulze:radio-4.0.4 to show that it works correctly with mechanize.

`disambiguate` broken because of unicode/str mix in `mechRepr` results

The following scenario error leads to an error:

A page contains a submit control whose value contains a non-ASCII character.
There is also a select with options or a text input field on this page.
browser.getControl() is used to select a not existing control.

This leads to a UnicodeDecodeError when zope.testbrowser tries to render the list of possible controls using the mechRepr of the controls. The exception happens when joining the mechRepr values of the controls because to a single string. The ones of the select options and text fields are unicode while the one of the submit control is str but contains a non-ASCII char.

Sphinx doctests are broken in Python 3

Maybe they even show that the code is broken because zope.testbrowser.cookies.expiration_string does not seem to work on Python 3, at least not the same way as in Python 2.

NotImplementedError in mechRepr breaks stuff

The NotImplementedError in https://github.com/zopefoundation/zope.testbrowser/blob/master/src/zope/testbrowser/browser.py#L769 breaks tests which expect a LookupError when there are controls that aren't covered in mechRepr. One example is an input element of type number for which the ctrl class is webtest.forms.Field. I'm not sure what the best fix here is. I'd say there should just be a fallback representation, but I'm not sure what it should look like. Maybe just as for webtest.forms.File?

Tests are broken on master

See https://travis-ci.org/zopefoundation/zope.testbrowser/builds/485980587

Cookie tests fail in different locales

tox -e docs fails for me. They pass for me (given #64) if I set LC_TIME=C instead of my usual lt_LT.UTF-8.

I suspect expires.strftime('%a, %d %b %Y %H:%M:%S GMT') in zope.testbrowser.cookies is not really what we want to.

ItemControl.click() lets you uncheck radio buttons

Given a form like

    <input type="radio" name="selected" id="select-c1" value="c1" checked="checked" />
    <label for="select-c1">Label 1</label>
    <input type="radio" name="selected" id="select-c2" value="c2" />
    <label for="select-c2">Label 2</label>

if you do browser.getControl('Label 1').click(), it'll unselect the radio button.

This is not how browsers behave, and it does not match how the older mechanize-based zope.testbrowser used to behave. If you click on the selected radio button it should stay selected.

Prominently document backwards-incompatible changes in zope.testbrowser 5.0.0

The changelog for version 5.0.0 should clearly document which zope.testbrowser APIs have disappeared, and what users are supposed to do when they upgrade their zope.testbrowser.

getLink uses stale cached HTML after goBack

The following pseudocode sequence goes wrong with zope.testbrowser 5.4.0:

browser.open(...)
browser.getLink(...).click()
browser.title
browser.goBack()
browser.getLink(...)

The second call to getLink incorrectly uses HTML from the previous response, cached by evaluating browser.title.

Browser._html and other related caches are invalidated by the Browser._changed method, but this is only called from Browser._preparedRequest; just navigating through history doesn't touch them. I think it would make a lot more sense to do the invalidation in Browser._setResponse, since most of the cached properties are derived from the response.

Browser._preparedRequest is a bit overenthusiastic about setting Referer

mechanize used to have this comment in its Referer header handling:

        # See RFC 2616 14.36.  The only times we know the source of the
        # request URI has a URI associated with it are redirect, and
        # Browser.click() / Browser.submit() / Browser.follow_link().
        # Otherwise, it's the user's job to add any Referer header before
        # .open()ing.

By contrast, if self.url is set at the start of Browser._preparedRequest, it just sets Referer to that. This means that if you do browser.open() twice in a row, the second one gets its Referer set to the first URL.

mechanize's behaviour was clearly more correct here. The RFC it quotes says: "The Referer field MUST NOT be sent if the Request-URI was obtained from a source that does not have its own URI, such as input from the user keyboard." In the context of a test browser, browser.open is analogous to input from the user keyboard and should not result in sending a Referer header.

I ran into some very confusing test failures in Launchpad that I eventually tracked down to this. I hacked around it by doing browser._response = None just before the relevant browser.open() calls.

displayValue setter silently ignores values which are not in displayOptions

Given I have a form:

<form>
  <select name="a">
    <option value="b" selected="selected">b</option>
 </select>
</form>

Trying to select something which is not an option using:

browser.getControl(name="a").displayValue = ['c']

silently ignores the unknown values. This results in no selected options in the example above.

I think in version 4.x an error was raised because there is no displayOption with the required title. (But I currently do not have code at hand to check this as 4.x calls into mechanize to select the value.)
The current behaviour is error prone for typos.

I propose to raise an exception (like KeyError) in this case.

Any suggestions?

ItemControl.selected = True fails if selected is already True

Given a form like

            <select multiple name="select-multiple">
              <option value="foo" />
              <option value="bar" />
              <option value="baz" />
            </select>

if you find the ItemControl for foo (call it item) and set item.selected = True twice in a row, zope.testbrowser ends up calling WebTest's MultipleSelect.value__set() with a list of values that contains duplicates, which in turn causes WebTest to raise

ValueError: Option(s) u'foo' not found (from u'foo', u'bar', u'baz')

Browser._findAllControls() has wrong behaviour with no-name inputs and checkboxes

Checkboxes become multiple controls that are handled together
when they are collected by Browser._findAllControls(). Usually they all
have the same name. When there are other fields with the same
name – or in a more normal and case multiple inputs with no name – unexpected errors are possible.
Has anybody any idea how to fix this?
Those tests assert the expected behaviour but fail with the current implementation.

class TestsSameNameAndCheckboxes(unittest.TestCase):
    """Testing weird behaviour regarding checkboxes with the same name."""

    def test_checkbox_controls_with_same_name_1(self):
        """A checkbox without name doesn't interfere with
        throwing a LookupError. In this case, multiple input
        fields are handled together but the checkbox is not
        treated as that because the first input is a button.
        """
        app = QuietTestApp()
        browser = Browser(wsgi_app=app)
        app.set_next_response(b'''\
<html><body>
<form action="." method="post" enctype="multipart/form-data">

    <button type="button">Do Stuff</button>
    <button type="button">Do Other Stuff</button>
    <label for="checkbox">Label</label>
    <input type="checkbox" id="checkbox">
</form></body></html>
        ''')
        browser.open('http://localhost/')
        with self.assertRaises(LookupError):
            browser.getControl('Not There')

    def test_checkbox_controls_with_same_name_2(self):
        """A checkbox is regognized as an ItemControl."""
        app = QuietTestApp()
        browser = Browser(wsgi_app=app)
        app.set_next_response(b'''\
<html><body>
<form action="." method="post" enctype="multipart/form-data">

    <button name="any" type="button">Do Stuff</button>
    <button name="any" type="button">Do Other Stuff</button>
    <label for="checkbox">Label</label>
    <input name="any" type="checkbox" id="checkbox">
</form></body></html>
        ''')
        browser.open('http://localhost/')
        control = browser.getControl('Label')
        assert isinstance(control, ItemControl)

    def test_checkbox_controls_with_same_name_3(self):
        """A checkbox without name doesn't interfere with
        throwing a LookupError. This time it seems to interpret the buttons
        as options of the checkbox."""
        app = QuietTestApp()
        browser = Browser(wsgi_app=app)
        app.set_next_response(b'''\
<html><body>
<form action="." method="post" enctype="multipart/form-data">
    <label for="checkbox">Label</label>
    <input type="checkbox" id="checkbox">
    <button type="button">Do Stuff</button>
    <button type="button">Do Other Stuff</button>
</form></body></html>
        ''')
        browser.open('http://localhost/')
        with self.assertRaises(LookupError):
            browser.getControl('Not There')

Replace `.browser.SetattrErrorsMixin` with `slots`?

.browser.SetattrErrorsMixin prevents spelling errors when using certain classes in this packages. It maybe pre-date __slots__ (Python 2.2) which can be used to achieve the same goal.

on the wire tests fails

The over_the_wire.txt test fails on the initial request too not only on the later request with a query string. Probably google changed the rules.

Probably we should skip/remove the on the wire tests at all because it's not what the testbrowser is used to do.

D:\git\zope\zope.testbrowser>bin\test.exe -pv1
Running tests at level 1
Running zope.testrunner.layer.UnitTests tests:
  Set up zope.testrunner.layer.UnitTests in 0.000 seconds.
  Running:
    29/42 (69.0%) ... e\zope.testbrowser\src\zope\testbrowser\over_the_wire.txt

Failure in test d:\git\zope\zope.testbrowser\src\zope\testbrowser\over_the_wire.txt
Failed doctest test for over_the_wire.txt
  File "d:\git\zope\zope.testbrowser\src\zope\testbrowser\over_the_wire.txt", line 0

----------------------------------------------------------------------
File "d:\git\zope\zope.testbrowser\src\zope\testbrowser\over_the_wire.txt", line 21, in over_the_wire.txt
Failed example:
    browser.open('http://google.com/ncr')
Exception raised:
    Traceback (most recent call last):
      File "C:\Python27\lib\doctest.py", line 1289, in __run
        compileflags, 1) in test.globs
      File "<doctest over_the_wire.txt[3]>", line 1, in <module>
        browser.open('http://google.com/ncr')
      File "d:\git\zope\zope.testbrowser\src\zope\testbrowser\browser.py", line 244, in open
        self._processRequest(url, make_request)
      File "d:\git\zope\zope.testbrowser\src\zope\testbrowser\browser.py", line 266, in _processRequest
        resp = make_request(reqargs)
      File "d:\git\zope\zope.testbrowser\src\zope\testbrowser\browser.py", line 242, in <lambda>
        make_request = lambda args: self.testapp.get(url, **args)
      File "d:\home\.buildout\eggs\webtest-2.0.14-py2.7.egg\webtest\app.py", line 212, in get
        expect_errors=expect_errors)
      File "d:\git\zope\zope.testbrowser\src\zope\testbrowser\browser.py", line 81, in do_request
        self._assertAllowed(req.url)
      File "d:\git\zope\zope.testbrowser\src\zope\testbrowser\browser.py", line 78, in _assertAllowed
        raise RobotExclusionError(url, 403, msg, [], None)
    RobotExclusionError: HTTP Error 403: request disallowed by robots.txt


  Ran 42 tests with 1 failures, 0 errors and 0 skipped in 1.626 seconds.
Tearing down left over layers:
  Tear down zope.testrunner.layer.UnitTests in 0.000 seconds.

Tests with failures:
   d:\git\zope\zope.testbrowser\src\zope\testbrowser\over_the_wire.txt

add support for setting request headers (user-agent)

Would be nice if there was support in zope.testbrowser.

See for example https://mail.zope.org/pipermail/zope3-users/2008-May/007871.html

5.0.0 Release available?

When will a release be available for the new wsgi app based 5.0.0 version? I'm asking because It looks very promising.

Tests failing on winbot

There's one test failure caused by webtest trying to guess the mimetype of the uploaded file because, apparently, Windows has never heard about CSV files.

This can be fixed by explicitly passing the MIME type to webtest.forms.Upload(filename, body, mimetype). This requires WebTest >= 2.0.8.

See Pylons/webtest#86

getLink() used to consider the whole enclosed text content, now it only looks at the directly contained text

Here's another of those mechanize-to-webtest compatibility issues:

<a href="..."><span>My link</span></a> used to be findable with browser.getLink('My link'), since mechanize collects all nested text, but now it's not found anymore, since zope.testbrowser only looks at the directly contained text.

getLink() used to see iframes, now it doesn't

The old mechanize-based zope.testbrowser allowed the following:

>>> print(browser.contents)
 ...
  <iframe width="..." height="..." src="http://localhost/style_preview.html"></iframe>
...
>>> browser.getLink(url='preview')
<Link text=None url='http://localhost/pdr/style_preview.html'>
>>> browser.getLink(url='preview').click()

The new WebTest-based zope.testbrowser does not treat iframes as links.

(I'm not arguing that it should, but maybe we want to add a note about this to the documentation and WONTFIX this bug?)

Allow option to prevent auto redirection

I saw there always following Location URL,
https://github.com/zopefoundation/zope.testbrowser/blob/master/src/zope/testbrowser/browser.py#L278

but previously in an older version, I used to browser.mech_browser.set_handle_redirect(False)

Is there any way to do something like this?

Travis thinks: "This is not an active repository"

See https://travis-ci.org/zopefoundation/zope.testbrowser on the tab named "current"

Someone with admin permissions in this repository should be able to fix this.

Fix DeprecationWarnings

e.g.,

zope/testbrowser/_compat.py:44: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working
  from collections import MutableMapping  # noqa