~alan-griffiths/mir/fix-server-crash : changes

~alan-griffiths/mir/fix-server-crash » Changes from revision 3732

From Revision 3732 to 3713

Rev	Summary	Authors	Date
3732	Tidy headers Tidy headers	Alan Griffiths	7 years ago
3731	A failure to draw one renerable should not crash the server A failure to draw one renerable should not crash the server	Alan Griffiths	7 years ago
3730	Add support for custom parameters in eglapp, presently for t... Add support for custom parameters in eglapp, presently for the camera apps to choose their capture device. . Approved by mir-ci-bot, Cemil Azizoglu.	Daniel van Vugt	7 years ago
3729	eglapp: Log valid DPI numbers (LP: #1559831) and more. eglapp: Log valid DPI numbers (LP: #1559831) and more. . Fixes: https://bugs.launchpad.net/bugs/1559831. Approved by mir-ci-bot, Alexandros Frantzis.	Daniel van Vugt	7 years ago
3728	Loosen performance expectations for: Loosen performance expectations for: TEST_F(TestClientInput, receives_one_touch_event_per_frame) The test itself remains fairly strong still. We emit 8.3 touches per frame, expecting that to become 1 touch per frame but allow up to 1.8 touches per frame before it's a fail. Still in the right ballpark. Judging by the last set of failures on Brandon's branch this should be sufficient to get a pass. (LP: #1570698). Fixes: https://bugs.launchpad.net/bugs/1570698. Approved by mir-ci-bot, Brandon Schaefer, Cemil Azizoglu.	Daniel van Vugt	7 years ago
3727	remove mir_buffer_get_native_buffer() from the gated MirBuff... remove mir_buffer_get_native_buffer() from the gated MirBuffer api. Given that we're defining our own platform (MirRenderSurface), it doesn't make sense to have a function that gives access to the underlying platform's native type (as we're saying there are no more underlying platforms anymore). Also define the MirNativeFence to simply be a Fd. Approved by mir-ci-bot, Alexandros Frantzis, Cemil Azizoglu.	Kevin DuBois	7 years ago
3726	debian: Remove unnecessary cmake-data build-dep debian: Remove unnecessary cmake-data build-dep cmake depends on cmake-data anyway, and removing the build-dep helps resolve cross-building issues with recent sbuild versions. Fixes: https://bugs.launchpad.net/bugs/1628507. Approved by mir-ci-bot, Cemil Azizoglu.	Alexandros Frantzis	7 years ago
3725	ipc: add the ability to pass MirBuffers in the nested server... ipc: add the ability to pass MirBuffers in the nested server to the nested clients. This needed some rework of android's fence-passing mechanism to be a bit more generic, so the nested platform can forward fences along properly. Approved by mir-ci-bot, Brandon Schaefer, Cemil Azizoglu.	Kevin DuBois	7 years ago
3724	Handle exceptions thrown by Display::configure() Handle exceptions thrown by Display::configure() We don't pre-validate display configurations, and I'm not sure it's possible in general to do so, so we need to handle errors in Display::configure() in a way that doesn't bring down the server. This sends an asynchronous error to clients when their display configuration fails to apply. It fails to handle problems where the shell-set configuration fails; that will be handled in a followup branch. Approved by mir-ci-bot, Kevin DuBois, Cemil Azizoglu.	Christopher James Ha...	7 years ago
3723	tools: Update check for publishable file locations. tools: Update check for publishable file locations. Approved by Kevin DuBois, Cemil Azizoglu, mir-ci-bot.	Alexandros Frantzis	7 years ago
3722	src/common/symbols.map: Remove some duplicate entries src/common/symbols.map: Remove some duplicate entries No ABI change from this. Just removing pure duplicates. If only our tooling was smarter and noticed these as well as other mistakes. Approved by Cemil Azizoglu, Alexandros Frantzis, mir-ci-bot.	Daniel van Vugt	7 years ago
3721	remove the --nbuffers option from the serverside and remove ... remove the --nbuffers option from the serverside and remove the BufferQueue system that was backing the server-side buffer management. MIR_CLIENT_NBUFFERS is available for tweaking the nbuffers option for bufferstream (management has moved client-side by design). Approved by mir-ci-bot, Alexandros Frantzis, Daniel van Vugt.	Kevin DuBois	7 years ago
3720	scene: Break potentially infinite loop in msh::AbstractShell... scene: Break potentially infinite loop in msh::AbstractShell::focus_next_session() (LP: #1625401). Fixes: https://bugs.launchpad.net/bugs/1625401. Approved by mir-ci-bot, Alexandros Frantzis, Daniel van Vugt.	Alan Griffiths	7 years ago
3719	mir_presentation_chain api: mir_presentation_chain api: fix situation where where the server would not notify a client of a buffers return, following the destruction of the MirPresentationChain that the buffer was associated with. fixes: LP: #1626503. Fixes: https://bugs.launchpad.net/bugs/1626503. Approved by Alexandros Frantzis, Brandon Schaefer, mir-ci-bot.	Kevin DuBois	7 years ago
3718	Allow changing the keymap used for a device during runtime. ... Allow changing the keymap used for a device during runtime. Fixes: https://bugs.launchpad.net/bugs/1626435. Approved by mir-ci-bot, Daniel van Vugt, Kevin DuBois, Brandon Schaefer.	Andreas Pokorny	7 years ago
3717	One definition for GammaCurve. One definition for GammaCurve. Approved by Kevin DuBois, mir-ci-bot.	Daniel van Vugt	7 years ago
3716	Fix ABI break LP: #1617865 without introducing another break... Fix ABI break LP: #1617865 without introducing another break by re-exporting the old deleted symbol versions. The new versions continue to also be exported by the usual symbols.map. . Fixes: https://bugs.launchpad.net/bugs/1617865. Approved by mir-ci-bot, Kevin DuBois, Alan Griffiths.	Daniel van Vugt	7 years ago
3715	add a private helper that the nested server can use to glean... add a private helper that the nested server can use to glean the parameters to use for eglCreateImage() so that MirBuffers can be uploaded as textures. We have the current hole in the MirBufferStream and MirBuffer interfaces that uploading a buffer as a texture is difficult and platform dependent. (see below for platform considerations). We have the future hole in the define-our-own-platform (a.k.a. MirRenderSurface) that we're not sure how we'll support eglCreateImage for our platform. (which just needs some more discussion) The nested server needs a way to upload a MirBuffer now for the nested-passthrough work. This MP provides the functionality now via this private helper function (so the nested platform doesn't have to know the android/mesa/eglstreams details). Once the MirRenderSurface transition is complete, we can just use the standard way to do this for the "mir egl platform", and remove this helper code. Platform considerations: On android we should be using ANDROID_image_native_buffer, and we have to get the actual ANativeBuffer pointer, so that refcounting on that object can work correctly. Some android platforms might work with KHR_image_pixmap, but this is tagged as legacy in the android code base (https://android.googlesource.com/platform/frameworks/native/+/master/include/ui/ANativeObjectBase.h#28), so I wouldn't expect it to work. On mesa, we can use KHR_image_pixmap or EXT_image_dma_buf_import, with the latter being more flexible. These need the gbm_bo or the prime_fd, and we have to bind MIR_SERVER_EGL_OPENGL_API to use them. Approved by mir-ci-bot, Brandon Schaefer, Alexandros Frantzis.	Kevin DuBois	7 years ago
3714	Improve the robustness of acceptance test: Improve the robustness of acceptance test: TestClientInput.receives_one_touch_event_per_frame so that it is now less likely to fail on a heavily loaded or slow system. (LP: #1570698) . Fixes: https://bugs.launchpad.net/bugs/1570698. Approved by mir-ci-bot, Chris Halse Rogers.	Daniel van Vugt	7 years ago
3713	Fix a couple of recent mistakes that are making 'make abi-ch... Fix a couple of recent mistakes that are making 'make abi-check' fail to build: * Missing header guard (LP: #1626866) * Include paths no longer exist (they moved but seemingly also unused). Fixes: https://bugs.launchpad.net/bugs/1626866. Approved by Brandon Schaefer, mir-ci-bot.	Daniel van Vugt	7 years ago

Older »