Prior to this commit, a benchmark test could report 0..n passes and 0..m
fails or skips, where n is the number of accumulation iterations used to
collect benchmark data and m is the number of times the test function
was invoked. Depending on the type of benchmark measurer being used,
this could result in a very large volume of test output and inconsistent
pass, fail and skip counts between test runs.
This commit changes the behaviour so that each benchmark test reports
one pass, fail or skip, regardless of the number of iterations used to
collect benchmark data.
This commit also prevents benchmark data being reported in the test
output if the benchmark test failed or skipped, as any benchmark data is
of dubious value in such cases.
The latter change in behaviour requires a minor modification to the
badxml selftest, which now tests quoting of literal strings in xml test
output for both passing and failing benchmarks.
Finally, this commit also adds a new selftest specifically for verifying
correct behaviour for benchmarks that fail or skip.
Task-number: QTBUG-24313
Change-Id: I3426dc659a7511b62fd183a031c7235bc753f497
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Prior to this commit, when an unexpected pass occurred the test output
showed the message "XPASS : tst_foo::function() 'expr' returned FALSE",
where the problem was actually that the expression evaluated to true
when it was expected to be false.
This commit changes the output to make it clear that the expression
evaluated to true unexpectedly.
Task-number: QTBUG-22118
Change-Id: Id22c178073d3b75789675ca37a8ef019029b1f91
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Prior to this commit, the following statement in the qtestlib
documentation was untrue: "If init() fails, the following testfunction
will not be executed, the test will proceed to the next testfunction."
If init() called QSKIP, the test function would be skipped, but if
init() reported a failure, the test function would still be executed
(even though doing so could be unsafe).
This commit makes testlib skip a test function if init() reports a
failure and enhances the selftests to cover skips and fails in both
init() and cleanup().
Task-number: QTBUG-20371
Change-Id: Id1cc8464ae0b8c257ae1b74dbe9189a501f5366b
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
For data-driven tests, testlib previously counted one fail or skip for
each data row that failed or skipped, while it counted only one pass
for a test function where all rows passed and counted no passes for a
test function where some rows passed and some rows failed. A similar
problem also existed for benchmark tests, which could run multiple
iterations of the same test, with each fail and skip being counted but
only a single pass being counted for the entire series of iterations.
This commit makes testlib count one pass, fail or skip for each data
row. Test functions that are not data-driven count one result for the
test function, as before. Benchmark tests count one pass, fail or skip
per iteration.
A side-effect of this change is that the test output in plain text, xml
and light xml formats now shows a result for every data row and
benchmark iteration executed, allowing post-processors to correctly
calculate the total number of tests executed. Previously, individual
rows were not shown in the test output if they passed, making such
calculations impossible.
The only change to the xunitxml output format is to correct a bug where
no test result was recorded for a test function if the last data row
was skipped and all other rows passed -- in which case the overall
result should be a pass. Note that there is also a pre-existing bug
in the xunit logger, where no result is reported if all rows are
skipped; that bug is unaffected by this commit.
Task-number: QTBUG-21848
Task-number: QTBUG-22124
Change-Id: I7e17177e10d6e89e55b9684c159bd506f21d002b
Reviewed-by: Ed Baak <ed.baak@nokia.com>
Set QT_TESTCOCOON_ACTIVE environment variable when the coverage is installed
for a test and unset it when the coverage data is saved. Tests that run when
QT_TESTCOCOON_ACTIVE is set are subtests and will not be considered as
stand-alone tests for the coverage.
When a test is run as a subtest its coverage data will not be saved for
itself but for the main test it is merged with. Also its status will not be
reported since only the status of the main test is expected in the test report,
e.g. the test tests/auto/testlib/selftests.
Change-Id: Icfdf99300aae18040e1a3441a8af21f68df4c0db
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Correct a bug in the test and use unique names for data rows. The bug
was that the test assumed that QCOMPARE for float values 100001 and
100002 would fail, but it actually succeeds. QCOMPARE for floats uses
qFuzzyCompare(), which succeeds if the numbers differ by no more than
1/100,000th of the smaller value. Thus QCOMPARE(100001, 100002) passes,
while QCOMPARE(99998, 99999) fails.
Change-Id: Ia35d3126c2e3ebe91d64daa309048514a365d9fb
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Some of the subtests are only run with plain-text output format. For
those subtests, the other output formats were unused and gradually
becoming out-of-date.
Change-Id: I4c10f7f5bab2d2cc7d2d2ad641fbf5d4df02b798
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Our OSX 10.7 test machines send a SIGILL rather than a SIGSEGV to a
process which attempts to dereference a null pointer.
Change the "crashes" test to dereference an invalid pointer with a
value slightly greater than 0 so that we get the same crash behavior on
all (unix) platforms.
Change-Id: I700a2c7d654a9468af5e5996010a258695ed2ae5
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Previously, ignored messages were only cleared at the end of each test
function, i.e. after all data rows were finished. This meant that if a
data row in a data-driven test function didn't cause all of the expected
messages to be generated, the remaining messages would be carried over
to the next data row. This would result in errors about missing
messages being associated with the last data row rather than with the
correct data row.
This commit makes testlib check for missing ignored messages after
running each data row rather than only doing so after the last data
row.
This commit also adds a regression test to demonstrate that ignored
messages can no longer be carried over from one data row to another.
Change-Id: Ibee51aa6e96866fbcbcb4acee1a8340a86a6a4ba
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Previously, expected failures were only cleared at the end of each test
function, i.e. after all data rows were finished. This meant that if a
data-driven test function called QEXPECT_FAIL and then didn't perform
any further verification steps to trigger the expected failure, the
expected failure would be carried over to the next data row, probably
causing the first verification step in the test function to XPASS (with
a seemingly irrelevant error message) for the next data row.
This commit adds the new function QTestResult::finishedCurrentTestData()
to cleanup after each data row is executed. This function treats calls
to QEXPECT_FAIL without subsequent verification steps as a test failure.
This commit also adds a regression test to demonstrate that expected
failures can no longer be carried over from one data row to another. If
run against the previous version of testlib, the new test would report a
pass instead of an error.
Change-Id: Ida5c7f080815b0dca9531131fed582b0918334cb
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
- Needed to add toLower() for compare as c:\ and C:\ was causing
a failure.
Change-Id: Idd55774d118b7249e88362688166b4f9d9712c7c
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Rename some of the test functions to improve consistency, and change the
order so that all the xfail tests are together (some new xpass tests
will be added later).
Change-Id: Ice1ac5bf61bc26109bb0d08f72eb74e1a0424101
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
The test simply generates each possible pair of Pass, Fail and Skip
results. At present the test simply serves to demonstrate the current
shortcomings of testlib's plain text logging, namely:
* If a test function passes for all data rows, that is counted as one
pass, but each skipped or failed row counts as one skip or fail.
* Only skipped and failed rows are reported individually in the test
output. Passed rows are not reported, so it is impossible to see how
many rows were executed.
* A skip followed by a pass will be reported as an overall pass for the
test function, but the same rows in reverse order will not report any
overall result for the test function.
Future commits will attempt to correct these problems.
Task-number: QTBUG-22124
Change-Id: If8c7ea15fc43ba9a1bccd0e881c1efc18e705b25
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
As in the past, to avoid rewriting various autotests that contain
line-number information, an extra blank line has been inserted at the
end of the license text to ensure that this commit does not change the
total number of lines in the license header.
Change-Id: I311e001373776812699d6efc045b5f742890c689
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Add a test case for calling QEXPECT_FAIL in a data-driven test without
passing a specific data tag for the expected failure, in which the
expected failure should apply to the current data row.
Change-Id: I70a6592de77b1718c3a749fd2216e66f8f312805
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Add a test for the case where a call to QEXPECT_FAIL nominates a
different data row from the current row, in which case the call to
QEXPECT_FAIL should be ignored.
Change-Id: Iff774b89f9374fa26105acdd8769177272a95229
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Add a test for the case where QEXPECT_FAIL is called twice in a row
without an intervening verification step, which is supposed to result in
a failure.
Change-Id: Ie004652a76a2c35c2679945670ff5561e8f5b62e
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Make the test fail if calling QEXPECT_FAIL with the mode set to Abort
does not cause the current test function to abort. Previously, such an
error would not cause the selftest to fail and someone would have to see
the warning message in the test output to know that there is a problem.
Change-Id: I2bd58fafe6b51ea0ab86fde7c5afb781e4534ee4
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
The existing expectfail selftest did not test QEXPECT_FAIL with a
data-driven test function. This commit adds such a test.
Change-Id: I39fa9aa227b58779ce5268dd37bf55468e7269c5
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
This test relies on being able to set a custom QLibraryInfo::PrefixPath
for the duration of the test. However, the test code neglected to
account for the following behavior on mac (from "Using qt.conf"
documentation):
On Mac OS X, the Prefix is relative to the Contents in the application
bundle.
Change-Id: Ie9b6d5ebfe8af7d7f332e827069b60a830d6c6f2
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
- Changed to detect valgrind at runtime and skip test instead of failing.
- subprograms inherit QT_QPA_PLATFORM value from parent if set.
Change-Id: I280acee389df1ee74ee6758a0dd1601226e103c7
Reviewed-by: Kurt Korbatits <kurt.korbatits@nokia.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
- All subprograms installed as part of test instead of separate installs
- Added installTestHelperApp() to testcase.prf
Change-Id: I02fbbb299f095054c9216ad0e5f91f574fb0fe3d
Reviewed-by: Kurt Korbatits <kurt.korbatits@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
The previous change missed some headers from years prior to 2011, and a
few new files were merged after the previous change.
Change-Id: Ib7d1a2b7062228c2a5373da64242b2ee1f0981e1
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Changed selftests unittest to use specific classes of QtCore instead of
pulling in all of QtCore headers by using include <QtCore>
- Decreasing build time.
Change-Id: Ic53a70423ecafeb342cc05ae6d7dee795533d205
Reviewed-by: Kurt Korbatits <kurt.korbatits@nokia.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Changed these tests to use QFINDTESTDATA macro to detect location of testdata.
Checking for a specific file contained in the testdata so as not to be confused by empty directories
created during configure.
Change-Id: Iac2ac6304b6b9ac79e00886025b93ec0af5a8507
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
... and deprecate the old registerTimer() functions. The new pure-
virtual registerTimer() breaks source-compatibility. Subclasses cannot
be instantiated anymore, since the pure virtual function signature has
changed.
QAbstractEventDispatcher::TimerInfo is no longer a QPair. It is now a
struct with timerId, interval, and timerType members. This is a source
incompatibility that should only affect subclasses of
QAbstractEventDispatcher, which will need to pass 3 arguments to the
TimerInfo constructor instead of 2. If the subclass used QPair<int,int>
instead of the TimerInfo typedef, the QPair<int,int> declarations will
need to be replaced with TimerInfo.
Call the new registerTimer() function with the type from
QObject::startTimer(). Change all subclasses of QAbstractEventDispatcher
to reimplement the new virtual function. The type argument is unused at
the momemnt, except to ensure that registeredTimers() returns the type
each timer was registered with. Implementations for the various
dispatchers will be done in separate commits.
Author: Thiago Macieira <thiago.macieira@nokia.com>
Change-Id: Ia22697e0ab0847810c5d162ef473e0e5a17a904b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Introduce a specialization for qCompare(QImage,QImage) that
checks isNull, size and format and outputs verbose messages.
Check isNull, size similarly for QPixmap.
Add an autotest:
- Add test to cmptest and make it a GUI application since
QImage requires QGuiApplication.
- Make testlib/selftests capable of running X11-GUI applications
by passing DISPLAY.
- Ignore stderr output for cmptest
- Add test data
Change-Id: I2b29c7822fbeedf2b22c90889739ed7ff859ce92
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Required for launching the sub-processes. The test worked only
when launched locally, not from 'make check'.
Change-Id: I42c9202a7726c3135f94445fb336b2b8241535a4
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Changed to use QT_INSTALL_TESTS to install subtests.
Change-Id: I4a2a7bd2d3e7d6da34dbb922bf377bee98cdedb0
Reviewed-by: Kurt Korbatits <kurt.korbatits@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
The old test only verified the behaviour of QCOMPARE when comparing
lists that were different. Add data rows for comparing empty lists and
non-empty lists that are equal.
Note that testlib currently does not report passing data rows (only
failing rows and completely passing test functions), so the new data
rows do not cause any change in the expected test output.
Change-Id: I137650ce0ca6250cee36bd9cb74b01f8abd4e89c
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
The (disabled) alive selftest contained a test for the QStringList
specialization of the QTest::qCompare template. This test is unrelated
to the rest of the alive selftest, so move it to the cmptest selftest,
where QCOMPARE is tested.
Change-Id: Ic6f0e491dd3b3ce8b4ca1d49666a099815575eaa
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
The overload is used in Qt Creator
(see src/libs/extensionsystem/pluginmanager.cpp).
The use case here is an application whose internal
QObjects can be tested by passing a command line parameter.
For this use case, it is inconvenient to have to allocate
memory and create a char argv[]- array.
This reverts commit ad80d42f8e.
Change-Id: I2a2f91e2840100fd62743f6d03b33005d67b18f8
Reviewed-by: Daniel Teske <daniel.teske@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
- Do not run with empty environment. At least PATH
is required at least (Qt + MSVC/MinGW runtime)
- Account for MSVC's different formatting of double numbers.
Change-Id: Ic7b1cf4a16a88c5384347b2651b011ac13c92d70
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
The overload of QTest::qExec() that takes a QStringList is not used
anywhere in Qt's autotests, despite having been in the qtestlib API
since Qt 4.4.
This lack of use most likely derives from the fact that none of the
QTEST_MAIN macros use the overload, and more than 99% of Qt's tests
use those macros to avoid explicitly calling QTest::qExec().
Change-Id: I264b21d7fe1a9f2d565f748cf8bbe32414a73bb0
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
The subtests are intended to be executed by tst_selftests and therefore
don't need to have "CONFIG += testcase", which causes 'make check" to
treat them as stand-alone tests.
Change-Id: I1b47e0186be99c591bb1e03bbec6386518f8a449
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Make the test object names correspond to the test program names, so that
it is easier to see which test produced each block of test output.
Change-Id: Ib19a955938b549f27cdcaac83bfa9150285b396d
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
This test uses qt.conf to enforce a predictable tests install location.
This was not working in the shadow build case, as qt.conf was only
present in the source tree; embed it as a resource at :/qt/etc/qt.conf
so it works in the shadow build case too.
Change-Id: Iec708d9686c29a1fc5ff0020dc11a93a593d20c5
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Automated tests often need to load some data from external files.
Currently, a wide variety of approaches for this have been used in Qt
autotests, including:
- embed the source directory into the test binary at compile time, and
find the testdata relative to that; this fails when the source tree
is no longer available (e.g. when the tests are deployed to a device).
- use a path relative to the current working directory, and trust that
the caller always sets the current working directory such that the
testdata can be found; this fails when the caller uses a different
working directory than expected.
- use a path relative to QCoreApplication::applicationDirPath();
this fails when source tree != build tree (since testdata is not
automatically copied into the build tree).
- compile the files into the binary using the Qt resource system; this
should work, but does not allow for testing of code which genuinely
needs external files.
It seems that there is not a simple method for determining the testdata
path which can be reliably used in all circumstances, so various tests
have reinvented the testdata location method in different ways.
Therefore, this is a good candidate for an addition to the testlib API.
The current implementation of QFINDTESTDATA is able to find testdata
in all three of (build tree, install tree, source tree), in that order.
Change-Id: Ib2fed860723ccf437240da3b00db22dfe1a6b56c
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Each QVariant instance has internal storage which may be
used for well-know basic types. This patch changes the
behavior by delegating type dependent operation to QMetaType
class which knows more types than QVariant itself.
The patch significantly reduce amount of code in QVariant
implementation.
There are few side effects of this patch:
- better performance:
* for Core types when using Gui (QGuiVariant is able to
construct Core types)
* for small custom types (QVariant::Private::Data is used for all
types that has size small enough)
- comparing two QVariants can give different result for small custom
types (binary comparison instead of pointer comparison)
Change-Id: Ic17fa500d6a882110bfba896fd456c8e6c7a63a9
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
(Note: This feature is ported from Qt 4.8.
See the following commits:
01575deafb7d26ca2431374e92c6d71de96547c7
4866d1ba8afbab61e102942d1ea93b81fea053d6
)
Passing the -datatags option to a QTestLib program prints the
available data tags to standard output.
For completeness, the test case name is also printed
at the start of each output line. (Although the file name
is supposed to match the lower-case version of the test case
name, this is currently not true in all cases (particularly not
under tests/benchmarks/). Even if there was a script to enforce this
convention, the -datatags option provides this information in a
reliable way.)
Data tags for each test function (f() in this case) are printed in
four different ways depending on the presence of local and global
data tags:
Case 1: No tags:
tst_MyTestCasetst_MyTestCase f
Case 2: Local tags only:
tst_MyTestCase f local tag 1
tst_MyTestCase f local tag 2
...
Case 3: Global tags only:
tst_MyTestCase f __global__ global tag 1
tst_MyTestCase f __global__ global tag 2
...
Case 4: Local and global tags:
tst_MyTestCase f local tag 1 __global__ global tag 1
tst_MyTestCase f local tag 2 __global__ global tag 1
...
tst_MyTestCase f local tag 1 __global__ global tag 2
tst_MyTestCase f local tag 2 __global__ global tag 2
...
...
Note that the string __global__ is assumed to be highly unlikely to occur
in a data tag (if it does, an ambiguity results).
Change-Id: Ib51aa0c3c32ad52e52ce519729292cf8f0ec5d50
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Reviewed-by: Sergio Ahumada <sergio.ahumada@nokia.com>
