Adds cts test definitions to the testrunner/test_defs.xml.
Adds support for runtest --cts arg, which will run all cts tests.
This temporarily relies on the addition of a 'cts' attribute to the test
definition - a new xml format may be defined later that changes how cts tests
are identified.
This change is based on previous unsubmitted CL
https://android-git.corp.google.com/g/Gerrit#change,1702. The only delta wrt
to that change is the use of InstrumentationCtsTestRunner.
Update: rebased to latest donut.
Merge commit '6d73a98d6df871fbd7f19af61d84a2eb1d993b4b' into donut
* commit '6d73a98d6df871fbd7f19af61d84a2eb1d993b4b':
AI 149478: Reorganized the PDK TOC.
* changes:
Add docs/STABLE-APIS.TXT which contains the list of exposed frozen APIs in the NDK. Also update README.TXT and docs/ANDROID-MK.TXT accordingly, to document the use of LOCAL_LDLIBS.
This change is to adapt runtest to the InstrumentationTestRunner change which
dumps coverage data to the app's local data folder, instead of a fixed file on
the sdcard.
Also fixed adb_interace.StartInstrumentationForPackage, added better handling
when generating coverage for tests with unknown coverage targets, and reduced
the duplicate definitions of the "coverage_targets.xml" definition file.
* changes:
Force the use of 32-bit toolchain binaries on 64-bit systems. This is used to reduce the number of distributable prebuilt packages (i.e. linux-x86_64 is not needed anymore).
This is used to reduce the number of distributable prebuilt packages
(i.e. linux-x86_64 is not needed anymore).
Also fix build/tools/build-toolchain.sh to properly copy the GPL and LGPL
license files into the generated prebuilt packages.
Also add a .gitignore files to get less spew from 'git status'
Can't reproduce the NPE in ContentAssist listed in the
bug. However if things go really wrong there are a couple
objects that can be null so let's be defensive about them.
Its purpose is to contact the 'boot-properties' service during the
boot process and to set the list of system properties read from the
emulator program through it.
It must be launched from /system/etc/init.goldfish.rc as root.
It depends on the 'boot-properties' service being implemented by the
emulator (in a later patch). If not available, it will simply timeout.
This also fixes a potential crash in qemud which can happen in the following case:
- client sends a command to a service
- the service responds with one or more message
- the service immediately closes the connection before the client can read the data
what happened is that the FDHandler object was buffering out-going packets to the
client, but the Multiplexer acted on the disconnection immediately. This resulted
in:
- client_free() being called
- fdhandler_shutdown() being called
- the FDHandler being placed on the 'closing' list to be able to send
buffered packets later. And 'receiver_close' being called
- client_fd_close() being called, which would call client_free()
recursively
- the second client_free() call would free the object
- the first client_free() call would try to free the object again
- ==> dlmalloc() detects a double-free and immediately aborts.
the fixes simply avoids the calle to receiver_close() in fdhandler_shutdown()
since this function is called from an explicit shutdown request from the multiplexer,
it doesn't correspond to the case where the client has closed the connection itself.
