| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
Change-Id: I18be2b351eac574df73fef8efe8e6c1c48683370
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
|
|
|
|
| |
Change-Id: I9a199b4575cb98e4fd1cfcc220518c96e09629a3
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
| |
While the new cache architecture has a type and version field for the global
header, I totally forgot to add the same thing to mark the type and version
of the actual cached data.
There was a version field for Package/Intent/ApplicationInfo, but this was
badly implemented and could crash the system trying to load an incompatible
cache file, because the version check was done too late.
Change-Id: I755cc5101213b60996541ea7e872e3017550ec37
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
|
|
|
|
| |
Change-Id: Ie471b636fbfab8e0b20dd467f98f3df0cda8504a
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
|
|
|
|
|
|
| |
Code was broken by a merge from 5.13 (2887a8c)
Change-Id: I4f87bd07a047483f4fb3e6758186229dceabd1e0
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With the xdg shell, ping is not related to a particular surface.
WaylandCompositor associates the pong with its original WindowSurface
with a map of the serial number to WindowSurface.
If the WindowSurface is destroyed before pong arrives, the map will have
an invalid pointer leading to a compositor crash.
Change-Id: Ide32ac80d42bf8373e8180e12ab89b29e6d00341
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
(cherry picked from commit 7038e1681863c8a59c6c6a758fc7ec41f29f2be7)
Reviewed-by: Bernd Weimer <bernd.weimer@pelagicore.com>
|
|
|
|
|
| |
Change-Id: Idcce8c3972302b8621946a2ee2818aa8bf83a807
Reviewed-by: Bernd Weimer <bernd.weimer@pelagicore.com>
|
|
|
|
|
|
|
|
|
| |
The disable-installer build-time and run-time flags have not been supported
properly since the PackageManager rewrite.
Also removed the dependency to OpenSSL for "disable-installer" builds.
Change-Id: I56e76568eca570d722631d76702776ff6bb32e21
Reviewed-by: Bernd Weimer <bernd.weimer@pelagicore.com>
|
|
|
|
|
| |
Change-Id: Ic21556897edf547faff5ee78a4b9ff3aab10dafb
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
A SignalSpy in QML tests frequently calls processEvents(), which
unfortunately doesn't lead to a flush of the Wayland message queue.
A workaround is implemented that forces flushing by manually emitting
the aboutToBlock signal.
Task-number: AUTOSUITE-709
Change-Id: I1480d756d9eb00d16181c9c747a3f27bce88891d
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
|
|
|
|
|
| |
Task-number: AUTOSUITE-1541
Change-Id: I20ce306148125c9c993d46a2a0c0d82d587878cf
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
|
|
|
|
| |
Change-Id: I2ee4cc28fe1e71a76179255020ff94d75add478f
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The controller accepts the options for the command now also before
the actual command e.g. -ioe debug-application
This makes it easier for the user, as this is also the format printed
by QCommandLineParser::showHelp()
Showing the full help for all commands also when only the specific
help for a command should be provided, makes it very hard to see what
the correct usage for the command actually is.
We now show only a short help when a command is provided and the full
help in all other cases.
Change-Id: I6927cf465ba1dd46cfbf361af04a3bfc8d089b67
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
|
|
|
|
|
|
|
| |
Added support for legacy configuration options, ignored deprecated
ones and fixed caCertificates.
Change-Id: I1f3afd592646744f2786993a202bf51d4ba91385
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
The crash info sometimes didn't inclulde the QML backtrace, because
the signal stack was too small. This was probably hidden in the past
by two additional, but redundant stack allocations.
Also cleaned-up some code.
Change-Id: I4173a1687ae7f2548d2609b4a26074766dfd211d
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
|\
| |
| |
| | |
Change-Id: I6f0940de367b9b3901d05ee61d994a50c4f2ca8b
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The runtime could be in an indistinct state when the signal was
emitted. Now it is possible to shutdown the runtime in the signal
handler. Also fixed object lifetime in the in-process runtime.
Fixes: AUTOSUITE-1431
Change-Id: I2782f4d91872c1037462bbce171fa432ed43fb6e
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| | |
Change-Id: I0926e68e411c4285d70144b2db4b6a848b39a784
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
- The tst_main test failed in older appman versions, if there is a
leftover cache file from a newer version and vice versa
- Adapted test output to fit deferred logging output
- Added "--dbus" option values to bash completion
Change-Id: Ica2600c61992ab8e87683d468af3fbdec766f944
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
- Added two other signals, that generate a core dump by default to
the crash handler: SIGQUIT and SIGSYS.
- Sending SIGABORT to processes in the same group would generate a
core dump of another process - using SIGTERM now.
- Setup alternate signal stack only once and hence allocate the
memory only once (there can be only one alternate signal stack)
Change-Id: I861505668dc2484730ec79410612f55698b695c4
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
In-process "windows" need to wait for their QML onCompleted handlers
to be called before they can be added to the System UI to make sure
that their window properties are available. If multiple windows are
created, the order would depend on the existence of onCompleted
handlers. The order has been changed to always match the order of
their componentComplete() function calls. This happens (but is not
guaranteed) to be the same order as in multi-process mode.
Change-Id: I1431277f6fa160b1713e94fc27195b178377e98a
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| | |
Change-Id: I4589d9713f2ad0df2cbff0a95c0c74628c5fbf38
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| | |
Change-Id: I66b60afafa6824dc1f13569e5c9f173eb8ec5680
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When the binary is build in release mode, the debug information
are stripped and might not be very accurate.
Add a note to the stacktrace and add a note what can be done instead
e.g. using DYLD_IMAGE_SUFFIX on macos.
Change-Id: Iee1f05a52099ca691848a7d2d106eeceef896769
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Similar to windows, we always want to provide a good stacktrace
on mac and enable the libbacktrace support for release binaries
as well.
Change-Id: I1f7585c749505035e55fd44e088340a493b265d9
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Update to the latest version from: https://github.com/Gagi2k/libbacktrace
* Fix the symbol lookup to also work with release binaries
* Adding support for macos frameworks and libraries without dSYM
* Adapt macho.c to latest dwarf changes
* Merge remote-tracking branch 'rust/rust-snapshot-2018-05-22'
Change-Id: Ia985af4a2bf682d6b055d9209bf50c6f8d22d714
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
As windows is one of our development platforms, but not really
a target platform, enable the stack-walker also for release builds,
which is also what the installer provides.
Change-Id: Iecbf26cc4a6d7af7e996e77235ebba2744eceb1f
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Support for this is still there, but a pro-active deprecation warning
shouldn't hurt.
Change-Id: I6bde47d653f0a21653cb4beeca633617e4a54d06
Fixes: AUTOSUITE-1465
Reviewed-by: Nikolay Zamotaev <nzamotaev@luxoft.com>
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| | |
Fixes: AUTOSUITE-1490
Change-Id: I5b45f58ace718dbcc88554248dea816c40ef363b
Reviewed-by: Bernd Weimer <bernd.weimer@pelagicore.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
MingW ships with lower-case header names instead of the camel-case names that
MSVC is providing. This is not a problem when building on Windows, but
including a camel-cased header fails when doing a MingW cross-build on a
Linux machine with a case-sensitive filesystem.
Fixes: AUTOSUITE-1484
Change-Id: I2850e9febade90d292ee89af69325845eb4fb7b6
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
|\ \ |
|
| |\ \
| | |/
| | |
| | | |
Change-Id: I2a3c5ad6dd3c10076af1bb0fd84ef56d29d88c50
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
MingW ships with lower-case header names instead of the camel-case names that
MSVC is providing. This is not a problem when building on Windows, but
including a camel-cased header fails when doing a MingW cross-build on a
Linux machine with a case-sensitive filesystem.
Fixes: AUTOSUITE-1484
Change-Id: Iebfc9b58956eb274c3cccee070b4ea4f69e105a3
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This adds a Windows version to the already existing Unix mechanism.
Change-Id: I7d92699875730322507d19a26cce794918caf267
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Right now we cannot create a backtrace, but at least we can use
the crashhandler to kill all child processes and also print a
QML backtrace and other useful information
Change-Id: I865e4d33dce5f6e032f550cd680176e1b3b54e19
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This will handle all the usual suspects, but it will not be able to handle
SIGKILL.
Please note: this is Unix-only for now.
Change-Id: I1fcac4dd8ddf2ef12d988012b04c2cb74b6f5f35
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| | |
Don't quit right away, when the QmlEngine quits, but let the normal
shutdown procedure run. This prevents warnings "QProcess: Destroyed
while process is still running".
Change-Id: I2d96961052b71f05b8a51f59c7fdcc082342800e
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| | |
Change-Id: I212e8c26e1030302aa76c27c8b11ad682ee5a818
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| | |
Change-Id: Ia4b3688b2eb270905c37c726aa80474c4336de4b
Fixes: AUTOSUITE-1470
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| | |
Change-Id: I5d8e21c3cddb693bdda668b88b0920f53846c1bc
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Do not block other users to access P2P dbus to enable
running other appman-launcher-qml instances under different users
Enable tighter security via user namespaces
Change-Id: I6f16095aefe265cf89249bc42c02ec260cc03f57
Reviewed-by: Robert Griebl <robert.griebl@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Backtraces are not supported on MinGW at the moment, because StackWalker
doesn't compile due to a missing dbghelp import library.
Also improved the async-safety of output functions. On the Linux side, I
noticed a lot of dead-locks and crashes while generating the backtrace lately.
In addition, the getOutputInformation kitchen-sink function has been
split up a bit and moved to the logging.cpp where it logically belongs.
Change-Id: I354c0e4e212a247338faddb6e15d8d0831852086
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The bug was trickier than anticipated, because I thought that only the appman
side would check the meta-data. Turns out the controller tries to verify the
packageId as well, but the meta-data was just not sent over DBus. Why? Because
the Package object may not have been in model yet (on first installation), so
calling PackageManager::get(<id>) led to an empty (error) repsonse.
Also fixed a few left missing s/application/package/ replacements.
Change-Id: I6fc436270ad193203fca94e66e739ffd4baeae27
Fixes: AUTOSUITE-1461
Reviewed-by: Egor Nemtsev <enemtsev@luxoft.com>
|
| |
| |
| |
| |
| |
| | |
Change-Id: I61defc60c9d7616a67ee6c53af42f76de17be246
Fixes: AUTOSUITE-1461
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| | |
Change-Id: Iac90c09b8a117bdd96d162cec96584b5b4cb5298
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This way we can always generate up-to-date qmltypes when doing developer
builds on the desktop, while embedded builds (which cannot run the
dumpqmltypes tool during the build process) can just re-use the last
committed version.
Change-Id: I8adf058b7876ebd429a8e7ef264b7575d221d4fc
Fixes: AUTOSUITE-1252
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Also reworded it to be more specific if it becomes relevant.
Change-Id: I60a2f8fca5d41c5e9776148911fada84602426cd
Fixes: AUTOSUITE-1195
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| | |
Change-Id: Ibaf57c72bd2d2bd146c86aad1d41210f9d3d6ca2
Fixes: AUTOSUITE-1238
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Instead of requiring the boilerplate setup call to be copied into each and
every system-ui, we do the setup now implicitly in the Main constructor.
(Calling it explicitly beforehand doesn't hurt though)
Change-Id: I549288397d397330af1733f8982666c54b4724cd
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|
| |
| |
| |
| |
| |
| | |
Change-Id: I83fe51acffc18c68be7b8894c641d902fe922945
Fixes: AUTOSUITE-1217
Reviewed-by: Dominik Holland <dominik.holland@qt.io>
|