GitLab

Some of the old Dalvik tests have custom build rules - changed to
support new location of core.art in data/art-test.

Change-Id: I4a84c299e99c3567d8afe4f76e23a35ae26b754e

I'd have preferred to have a 512-byte limit, but there are some monsters
in the verifier; 2000-line functions and the like. I'm also not policing
tests (except for one silly one). They can use all the stack they like.

This fixes the IntMath test (the stack overflow test was failing because
we were using more than 4KiB to throw!).

Change-Id: I7e53e2fde2b39fde1910f8ee5b1712e8a66069c7

While addres (Dex PC) uses unsigned in advancing, line number uses
signed. Now 007-exceptions passes.

Change-Id: Ic65cd5093274c6b4d0e9949e5ea50efcdaf71322

Stress test the ReferenceMap when the program may have out-of-bound
exception.

Change-Id: Iec892d9564c501794571a7c52519ac539a792d6e

In old Dalvik, adjopcode was declared as int instead of uint8. And
it's declared in different scope.
Pass the art-target tests and art-host tests.

Change-Id: Ie65c16a43ca32f1acf425a740bebc02f005f9011

This rounds out the proxy implementation by adding missing pieces to the
class linker, extending tests and fixing issues in the runtime support.
There are also some tweaks for performance and to clean up Method/Object
a little.
A unit test of the functionality is "art/test/run-test 044"

Change-Id: Id94102d10b81cd9b12b95ba8618f6187490204c4

Restructured the type inference mechanism, added lots of DCHECKS,
bumped the default memory allocation size to reflect AOT
compilation and tweaked the bit vector manipulation routines
to be better at handling large sparse vectors (something the old
trace JIT didn't encounter enough to care).

With this CL, optimization is back on by default.  Should also see
a significant boost in compilation speed (~2x better for boot.oat).

Change-Id: Ifd134ef337be173a1be756bb9198b24c5b4936b3

Change-Id: I094375230af2d9a88e30245b390cac71be7b50f4

The key ingredient to this change is that the NULL check on the NewStringUTF
call in ClassLinker::FindClass was accidentally checking the input to the
function (which we know isn't null) rather than the output (which will be
if allocation fails).

I've also fixed a bunch of places where we were missing null checks on
allocation, but there's a lot more to do.

I've stopped -Xcheck:jni from trying to dump pending OutOfMemoryError stacks
because that's highly likely to go horribly wrong and just gets in the way
of debugging the actual problem. (We could perhaps do better, so I've added
a TODO.)

In DexFile, we were always allocating java.lang.Strings for the names of
each local variable, even if we were only parsing the debug info for the
purposes of translating a line number. This was less of a problem in Dalvik
where the names were just const char*s, but since we go straight to Strings,
this was causing yet more OOME carnage (and is a lot of wasted effort).

This fixes the regression we were seeing in 061.

Change-Id: I6241602756db3070b214ccb35e3760b37e1e3d10

This code compiles but has not been exercised. I still
need to wire in the code blocks and return types.

Change-Id: I1796103d3bc4fc3c863f95bdca4cfb9765d4b3b7

Change-Id: Ia84eadc085c2bd517d335af96c7ffaefd25af0fc

Change-Id: I7de0bb04adac4c68cb410cefec86f50bb30dddea

Change-Id: Ib4d641ca7000b0f563cb9ab9c0970d6b8ad92060

Change-Id: Ie9130eb0af4230bfe280d549c559762c7643e4ad

Change-Id: I9b48628e338a5346bc3ac21f3469e4e3d224a956

Also implement Zygote_nativeForkAndSpecialize

Change-Id: I7dfb257b2897279a4cdac4b1ca194c1ac84eb047

This gets us closer to passing 061 again, but the ClassLoader changes seem to
have made life a bit more recursive. (That is: this is not a complete fix.)

Change-Id: Ibd7e7a54a6cbfb17020295e034053e136777e891

oat tests, gtests, and run-test content to /data/art-test
boot.oat and boot.art to /data/art-cache

Change-Id: I68f9f6798bc4ea0596642ff5624f3e77dec6ccbe

Use CHECK, otherwise, test was wrongly shown to pass even when there is error.

Change-Id: I7c273db81c4ba0aaaa6a97ee1403e92ec6db6800

Change-Id: I719fd936992950f6d7039ca9cce7435ab1fec229

This also changes us over to one big stats update at the end of sweeping,
like Dalvik, rather than recording every single free individually.

Change-Id: Ib8e2a83d41d36e35e154183a4e173e915af79ae9

Make more aggressive use of the dex cache when we have it.

Change-Id: I125a7bac031f7f0cec68194b5380bdfd83c92833

Change-Id: I4561230971ed31ee9af0ec5007e0d09b30781779

Also remove the Thread::pthread_ field, which isn't really needed. We also
don't need to update the TLS pointer to our Thread*, because the
address will be the same in the child anyway.

Change-Id: I595832a682a03ccdbb510e8e9ce0a32d5c6526b3

Change-Id: I495f6e32873e7ea5ffde25490678b571312e34cd

Also changed -Xzygote to default boot.art and log arguments

Finally Makefile work:
- separate out test dependencies
- changed zygote with art to work by replacing libdvm with libart[d]
- changed Makefile defines to name and error check their arguments
  This was to fix bug in generating ART_TARGET_EXECUTABLES
  where we used $(1) when we meant $(3)

Change-Id: I7ef6008eff1fe99b59d151b6793e8fd8ea5304d4

Change-Id: I5d08ebe33048a1380d8f868e5e87c3aab7265770

Change-Id: I2b0cfdb8f0a66c1bac92f4c67a4be06fa78b6d74

Change-Id: I2f0c7b93e8ac3b3fa4bb350e787065f92c17972f

We _could_ fix this, but it would be ugly, potentially expensive, and it's
really not obvious there's any point. (At least until we have to worry about
DexFile*s not owned by the ClassLinker.)

Change-Id: I2404bfe15e36bba0a6d1d362a465f2fccf1223eb