Fix-365: Refactor TangoAttribute push_event #366

cpascual · 2017-01-13T12:01:43Z

Fix #365. (and also fix #261 as a side-effect)

Taurus4 introduced a regression related to the merge of TaurusAttribute
and TaurusConfiguration. The TangoAttribute.push_event method handles
all events but it doesn't do it properly for configuration-related
error events received before the TangoAttribute has been initialized.

In this situation, the TangoAttribute ends up flagged as subscribed
but the subscription was not actually done, preventing to become
properly subscribed when a non-error event comes later on.
It also may incorrectly deactivate the polling mechanism.

Refactor the push_event code to clean it and correctly handle conf
errors.

Notes:

As part of the review and integration of this change, we needed to deal
with the spurious failures in travis (Random failures in testsuite (workaround exists, but proper fix is pending) #261 ). This PR also makes the testsuite
more robust (not infalible, though). It is enough IMHO to Fix Random failures in testsuite (workaround exists, but proper fix is pending) #261.
this seems to solve the issue with the Sardana server reported in:
https://sourceforge.net/p/sardana/mailman/message/35591825/
(we tried with this change, with sardana from develop and the tango9 stack
from debian9 repos... and we successfully ran sar_demo using a single
Sardana instance)

Taurus4 introduced a regression related to the merge of TaurusAttribute and TaurusConfiguration. The TangoAttribute.push_event method handles all events but it doesn't do it properly for configuration-related error events received before the TangoAttribute has been initialized. In this situation, the TangoAttribute ends up flagged as subscribed but the subscription was not actually done, preventing to become properly subscribed when a non-error event comes later on. It also may incorrectly deactivate the polling mechanism. Refactor the push_event code to clean it and correctly handle conf errors.

cpascual · 2017-01-13T12:34:42Z

Note that travis shows many failures in tests... they need to be investigated before accepting this PR

cpascual · 2017-01-17T09:41:42Z

Investigating the travis errors, with @sagiss ... It seems like they are (again) linked to #261, i.e., they are spurious failures, hard to reproduce outside travis.

After forcing a re-run, we went from 8 errors to only 1:

======================================================================

/usr/local/lib/python2.7/dist-packages/taurus-4.0.3a0-py2.7.egg/taurus/core/resource/test/res/attr_resources_file.py

FAIL: test_texts (taurus.qt.qtgui.panel.test.test_taurusvalue.TaurusValueTest)

Testing texts(expected=('double_scalar', '1.23', '0.0 mm', 'mm'), model='tango:TangoSchemeTest/unittest/temp-1/double_scalar')

----------------------------------------------------------------------

Traceback (most recent call last):

  File "/usr/local/lib/python2.7/dist-packages/taurus-4.0.3a0-py2.7.egg/taurus/test/base.py", line 113, in newTest

    return helper(**helper_kwargs)

  File "/usr/local/lib/python2.7/dist-packages/taurus-4.0.3a0-py2.7.egg/taurus/qt/qtgui/panel/test/test_taurusvalue.py", line 77, in texts

    self.assertEqual(got, expected, msg)

AssertionError: wrong text for "tango:TangoSchemeTest/unittest/temp-1/double_scalar":

 expected: ('double_scalar', '1.23', '0.0 mm', 'mm')

 got: ('double_scalar', '-----', 'None', '')

----------------------------------------------------------------------

Ran 560 tests in 78.112s

FAILED (failures=1, skipped=11)

I am re-running it again in the hope that we get a clean build in travis to confirm that this is all #261 - related

mrosanes · 2017-01-17T09:49:14Z

On docker cpascual/taurus-test (launching the taurustestsuite outside travis), I managed to reproduce the failures that were appearing before on Travis.

root@taurus-test:/# 
...
======================================================================
FAIL: test_write_read_attr_13 (taurus.core.tango.test.test_tangoattribute.AttributeTestCase)
Testing write_read_attr(attrname='short_scalar', expected={'warnings': [, ], 'wvalue': , 'rvalue': , 'data_format': 0, 'label': 'short_scalar', 'writable': True, 'range': [, ], 'alarms': [, ], 'type': 0}, expected_attrv={'w_value': 1000, 'quality': , 'value': 1000}, setvalue=)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/taurus/lib/taurus/test/base.py", line 113, in newTest
    return helper(**helper_kwargs)
  File "/home/taurus/lib/taurus/core/tango/test/test_tangoattribute.py", line 771, in write_read_attr
    self.assertTrue(isinstance(read_value, TangoAttrValue), msg)
AssertionError: read() for "short_scalar" did not return a TangoAttrValue (got a NoneType)

======================================================================
FAIL: test_write_read_attr_16 (taurus.core.tango.test.test_tangoattribute.AttributeTestCase)
Testing write_read_attr(attrname='double_scalar', expected={'wvalue': , 'rvalue': , 'type': 1}, expected_attrv={'w_value': 3, 'quality': , 'value': 3}, setvalue=)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/taurus/lib/taurus/test/base.py", line 113, in newTest
    return helper(**helper_kwargs)
  File "/home/taurus/lib/taurus/core/tango/test/test_tangoattribute.py", line 771, in write_read_attr
    self.assertTrue(isinstance(read_value, TangoAttrValue), msg)
AssertionError: read() for "double_scalar" did not return a TangoAttrValue (got a NoneType)

======================================================================
FAIL: test_write_read_attr_17 (taurus.core.tango.test.test_tangoattribute.AttributeTestCase)
Testing write_read_attr(attrname='double_scalar', expected={'wvalue': , 'rvalue': }, expected_attrv={'wvalue': , 'rvalue': , 'quality': }, setvalue=)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/taurus/lib/taurus/test/base.py", line 113, in newTest
    return helper(**helper_kwargs)
  File "/home/taurus/lib/taurus/core/tango/test/test_tangoattribute.py", line 771, in write_read_attr
    self.assertTrue(isinstance(read_value, TangoAttrValue), msg)
AssertionError: read() for "double_scalar" did not return a TangoAttrValue (got a NoneType)

======================================================================
FAIL: test_write_read_attr_18 (taurus.core.tango.test.test_tangoattribute.AttributeTestCase)
Testing write_read_attr(attrname='double_scalar', expected={'alarms': [, ], 'range': [, ], 'name': 'double_scalar', 'warnings': [, ], 'wvalue': , 'rvalue': , 'type': 1}, expected_attrv={'wvalue': , 'w_value': 10, 'rvalue': , 'quality': , 'value': 10}, setvalue=)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/taurus/lib/taurus/test/base.py", line 113, in newTest
    return helper(**helper_kwargs)
  File "/home/taurus/lib/taurus/core/tango/test/test_tangoattribute.py", line 771, in write_read_attr
    self.assertTrue(isinstance(read_value, TangoAttrValue), msg)
AssertionError: read() for "double_scalar" did not return a TangoAttrValue (got a NoneType)

======================================================================
FAIL: test_write_read_attr_19 (taurus.core.tango.test.test_tangoattribute.AttributeTestCase)
Testing write_read_attr(attrname='float_scalar', expected={'wvalue': , 'rvalue': , 'type': 1}, expected_attrv={'w_value': 3, 'quality': , 'value': 3}, setvalue=)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/taurus/lib/taurus/test/base.py", line 113, in newTest
    return helper(**helper_kwargs)
  File "/home/taurus/lib/taurus/core/tango/test/test_tangoattribute.py", line 771, in write_read_attr
    self.assertTrue(isinstance(read_value, TangoAttrValue), msg)
AssertionError: read() for "float_scalar" did not return a TangoAttrValue (got a NoneType)

======================================================================
FAIL: test_write_read_attr_20 (taurus.core.tango.test.test_tangoattribute.AttributeTestCase)
Testing write_read_attr(attrname='float_scalar', expected={'wvalue': , 'rvalue': }, expected_attrv=': , 'rvalue': , 'quality': }, setvalue=)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/taurus/lib/taurus/test/base.py", line 113, in newTest
    return helper(**helper_kwargs)
  File "/home/taurus/lib/taurus/core/tango/test/test_tangoattribute.py", line 771, in write_read_attr
    self.assertTrue(isinstance(read_value, TangoAttrValue), msg)
AssertionError: read() for "float_scalar" did not return a TangoAttrValue (got a NoneType)

======================================================================
FAIL: test_write_read_attr_21 (taurus.core.tango.test.test_tangoattribute.AttributeTestCase)
Testing write_read_attr(attrname='float_scalar', expected={'alarms': [, ], 'range': [, ], 'name': 'float_scalar', 'warnings': [, ],': , 'rvalue': , 'type': 1}, expected_attrv={'wvalue': , 'w_value': 10, : , 'quality': , 'value': 10}, setvalue=)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/taurus/lib/taurus/test/base.py", line 113, in newTest
    return helper(**helper_kwargs)
  File "/home/taurus/lib/taurus/core/tango/test/test_tangoattribute.py", line 771, in write_read_attr
    self.assertTrue(isinstance(read_value, TangoAttrValue), msg)
AssertionError: read() for "float_scalar" did not return a TangoAttrValue (got a NoneType)

======================================================================
FAIL: test_write_read_attr_24 (taurus.core.tango.test.test_tangoattribute.AttributeTestCase)
Testing write_read_attr(attrname='boolean_scalar', expected={'writable': True, 'alarms': [None, None], 'range': [None, None], 'warnings': [None, None], ' False, 'rvalue': False, 'type': 3}, expected_attrv={'wvalue': False, 'w_value': False, 'rvalue': False, 'quality': , 'value': setvalue=False)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/taurus/lib/taurus/test/base.py", line 113, in newTest
    return helper(**helper_kwargs)
  File "/home/taurus/lib/taurus/core/tango/test/test_tangoattribute.py", line 771, in write_read_attr
    self.assertTrue(isinstance(read_value, TangoAttrValue), msg)
AssertionError: read() for "boolean_scalar" did not return a TangoAttrValue (got a NoneType)

----------------------------------------------------------------------
Ran 560 tests in 79.258s

FAILED (failures=8, skipped=11)

The test-timer test is not deterministic. It may fail occasionally (e.g. if the machine is overloaded) cause testsuite failures that interfere with CI. Exclude this test in travis.

AttributeTestCase.write_read_attr helper uses attr.write(with_read=True) for reading the value. This may lead to reading None in some cases (e.g. attr.isUsingEvents() returns True) and cause the test to fail. Do separated write and read (without cache) instead.

BaseWidgetTestCase based tests make calls to QApplication.processEvents to emulate the Qt main loop. It seems that a single call is sometimes insufficient to process all needed events for the test, which causes spurious failures (see taurus-org#261) As a workaround, implement a BaseWidgetTestCase.processEvents method that loops several times over QApplication.processEvents and use it in the tests. Also remove previous workaround based on sleeps after setModel. Also re-enable skipped tauruslabel background tests since this fix seems to make them work again.

Some tests were commented out because they break test isolation. The fix of taurus-org#365 may have solved the root cause of this, so reenable them.

Some random test failures (see taurus-org#261) seem to be related to test DS not being properly initialized. Add a waiting time (controlled by the TAURUS_STARTER_WAIT env var) during starting of DS to alleviate this. Set TAURUS_STARTER_WAIT to 5 in Travis (i.e., wait 5 s after each DS start)

cpascual · 2017-01-19T09:25:24Z

I included several commits (cherry-picked and reworked from cmft#3) that should boost the robustness of the testsuite (they fix/work-around causes of #261).

cpascual · 2017-01-19T09:50:31Z

@cmft and me reviewed (and pair-programmed) this, so I am accepting the PR

cpascual added the bug label Jan 13, 2017

cpascual modified the milestone: Jan17 Jan 13, 2017

cmft and others added 5 commits January 19, 2017 09:05

Skip test_timer in travis

14d1084

The test-timer test is not deterministic. It may fail occasionally (e.g. if the machine is overloaded) cause testsuite failures that interfere with CI. Exclude this test in travis.

Enable previously disabled tango attribute tests

c94e447

Some tests were commented out because they break test isolation. The fix of taurus-org#365 may have solved the root cause of this, so reenable them.

cpascual requested review from cmft and removed request for cmft January 19, 2017 09:45

cpascual merged commit 476ccca into taurus-org:develop Jan 19, 2017

cpascual mentioned this pull request May 22, 2017

Try again to work around race conditions in test suite #454

Merged

cpascual deleted the bug-365 branch July 14, 2017 13:23

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Fix-365: Refactor TangoAttribute push_event #366

Fix-365: Refactor TangoAttribute push_event #366

Uh oh!

cpascual commented Jan 13, 2017 •

edited

Loading

Uh oh!

cpascual commented Jan 13, 2017

Uh oh!

cpascual commented Jan 17, 2017

Uh oh!

mrosanes commented Jan 17, 2017

Uh oh!

cpascual commented Jan 19, 2017

Uh oh!

cpascual commented Jan 19, 2017

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Fix-365: Refactor TangoAttribute push_event #366

Fix-365: Refactor TangoAttribute push_event #366

Uh oh!

Conversation

cpascual commented Jan 13, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cpascual commented Jan 13, 2017

Uh oh!

cpascual commented Jan 17, 2017

Uh oh!

mrosanes commented Jan 17, 2017

Uh oh!

cpascual commented Jan 19, 2017

Uh oh!

cpascual commented Jan 19, 2017

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

cpascual commented Jan 13, 2017 •

edited

Loading