You could trigger invalid states by pressing the recording button
while the profiler was waiting for data to be loaded from the
application or by pressing the "stop" button twice. Now the buttons
are disabled when they don't make any sense. In particular you cannot
clear, change features, filter the display while recording and you
cannot toggle the recording while loading. Pressing the "stop" button
twice will cancel the loading now, similar to what happens if you
first disable recording and then stop while the data is loaded.
Change-Id: Ie5133f4bed76593dedc9763b778a2c5c17187543
Reviewed-by: Tobias Hunger <tobias.hunger@theqtcompany.com>
* Remove superfluous dot from test file name
* Reuse an existing member
Change-Id: I735dee09ae7ceef4509a4bda87118baaf4d11b01
Reviewed-by: Nikolai Kosjar <nikolai.kosjar@theqtcompany.com>
The move to LIBEXEC_PATH broke running the puppet from the packages on
Linux and OS X, because there it cannot find the platform plugin
anymore.
We need to add a qt.conf for it, and because Qt Creator.app/Contents/
Resources already contains the qt.conf for Qt Creator itself, the puppet
must also move to its own subdirectory on OS X.
Change-Id: I8e8cc48c49a01e10c7d7d3cc4c73020195cd5ca9
Reviewed-by: Thomas Hartmann <Thomas.Hartmann@digia.com>
You have seen always all setting but some settings had no meaning
because
only the foreground color is used. Now you can disable this settings so
it
is more clear what you can change.
Change-Id: I0fdd2ac6f40e27b5160a2c54a512289457674dae
Reviewed-by: David Schulz <david.schulz@theqtcompany.com>
* Set the category apart.
* Show only the disable option and only in debug mode.
Change-Id: If5f65f9bd04c8e37e376a9862035005f4456394a
Reviewed-by: Marco Bubke <marco.bubke@theqtcompany.com>
The third column was accidentally removed from one data row in
commit 271794fbc9
Fixes the following fatal error:
QFETCH: Requested testdata 'expectFailMessage' not available, check your _data function.
Change-Id: I4a09df2e044b4dcee1ed2ffc589082a3e2601232
Reviewed-by: Orgad Shaneh <orgads@gmail.com>
Reviewed-by: Christian Kandeler <christian.kandeler@theqtcompany.com>
It was all done in one huge, unreadable method.
The refactoring along-side fixes the following:
- iOS tool chains were demoted to manual at start up because there
was no tool chain factory that would auto detect them
- a QTC_CHECK failed because single Qt Versions could appear multiple
times in the architecture->Qt version map, because each architecture
is present in two abis (32-bit and 64-bit)
The refactoring also removes unneeded automatic conversion
from non-iOS Qt version to iOS Qt version, and removes considering GCC
toolchains (which were not handled anyhow, because only Clang
"platforms" were considered).
Change-Id: Ic9ae797646f159ed45959fc797990aa98f2136fb
Reviewed-by: Tobias Hunger <tobias.hunger@theqtcompany.com>
The recent optimization to avoid running toolchains during startup
has broken the registration of tool chains with the manager, resulting
in auto-detected toolchains being demoted to manual in one run and
then getting removed in the next (at which point a new set of toolchains
got auto-detected).
Change-Id: I175c7344238df1bcb36a8b4949f0f0faf8828116
Reviewed-by: Eike Ziller <eike.ziller@theqtcompany.com>
replace Q_ASSERT with QTC_ASSERT and be more consistent with asserts
Change-Id: I81161a1ee82218ef23f62cc90e52587c342e2e22
Reviewed-by: Orgad Shaneh <orgads@gmail.com>
Reviewed-by: Eike Ziller <eike.ziller@theqtcompany.com>
We are overstretching std::vector implementations in some
cases of partial self-assignment.
Task-number: QTCREATORBUG-15183
Change-Id: I144e9e34df117286a7eef6403e29054d530cacbe
Reviewed-by: Christian Stenger <christian.stenger@theqtcompany.com>
The original idea was that we would passively wait for another thread to
clean up the locker, hence the check-sleep-loop. This was all dandy,
except for *also* using the wait condition: this was a) mostly pointless
(it would just avoid a few iterations of the wait loop) and b) buggy (if
there were no other waiting threads, the actual reader thread wouldn't
know that it needs to wake somebody up).
As the passive waiting is ugly, we instead fix the use of the wait
condition, and do away with the loop.
Task-number: QTCREATORBUG-15181
Change-Id: I477dbe7cda49ceca9aa387910d94ad763a43012b
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Reviewed-by: Daniel Teske <daniel.teske@theqtcompany.com>
We need both clang.exe and clang-cl.exe for mingw and msvc
Change-Id: I2f8883f2ac42636699d4326d4d5f044b34b7d9ba
Reviewed-by: Eike Ziller <eike.ziller@theqtcompany.com>
Fix the following use case:
* Open foo.h
* Open foo.cpp
* Change both files externally (e.g. git checkout) e.g. by adding a
function (declaration + definition)
* Confirm the reload dialog in Qt Creator
* The added function in foo.cpp gets a diagnostic that it's out of line
although foo.h is also reloaded.
TextDocument::contentsChangedWithPosition() is also emitted during a
reload. However, the revision of the document at that point is not yet
updated. This led to an out-dated copy of the unsaved files in the clang
backend.
Change-Id: I70580d32bb7a34dc43356dc05019cc18addf950e
Reviewed-by: Marco Bubke <marco.bubke@theqtcompany.com>
Make the styling work better, especially when the kit information
is displayed in the MiniProjectTargetSelector.
Change-Id: I562c1c92942afbfec1b960740a326c4dd002864f
Reviewed-by: Daniel Teske <daniel.teske@theqtcompany.com>
Update the unit tests while at it. Add some new tests, rearrange
existing ones.
Change-Id: Icc2db644f8fe9752c1bf8e66b134738c27b0fb25
Reviewed-by: David Schulz <david.schulz@theqtcompany.com>
Show a nice spinner while waiting for data from Gerrit.
Change-Id: Ia35d7408e6a65126d40e8cff5278a442f4b5c760
Reviewed-by: Tobias Hunger <tobias.hunger@theqtcompany.com>
Same test as the first version, but without having the header open in
the editor. This one is also skipped since we can't make it reliable
without adding extra messages.
Follow up changes will improve the names of both tests and will make
them both reliable so that the can be activated again.
Change-Id: Ie3ca52e01aedd612bc07e7429dd64d92eb1a23b5
Reviewed-by: Marco Bubke <marco.bubke@theqtcompany.com>
...with the clang code model.
In updateProjectInfo() we lock the project mutex and emit a signal.
However, the correspondig slot in the clang code model will call back
into a function that needs to lock the mutex, too:
CppTools::CppModelManager::workingCopy // dead locks here
CppTools::BaseEditorDocumentParser::InMemoryInfo::InMemoryInfo
ClangCodeModel::Internal::ClangEditorDocumentProcessor::run
ClangCodeModel::Internal::ModelManagerSupportClang::unregisterTranslationUnitsWithProjectParts
ClangCodeModel::Internal::ModelManagerSupportClang::onProjectPartsRemoved
...
CppTools::CppModelManager::projectPartsRemoved
CppTools::CppModelManager::updateProjectInfo // emits here
Fixed by emitting the signal after releasing the mutex.
Change-Id: Id3f3cd22b0604f343ada56ea13c26f0dc1a75bd6
Reviewed-by: Marco Bubke <marco.bubke@theqtcompany.com>
