GitLab

This prevents the generation of a new dexfile when an odex file already
exists when switching between Art and Dalvik.

Bug: 13463154
Change-Id: I55ef9abe1fe172ee656550e78556381f95698c33

Bug: 11519714

The bug was due to the fact that when merging 2 dexes, not enough size may be reserved for the "typeLists" section.
This is because we only aligned the sum of the size of both sections.
If both typeLists sections were aligned with 2 and not with 4, we are 4 bytes short when writing.

Change-Id: I73b51eb25434a622143011741a69b88d42507f43
(cherry picked from commit 2241dbe1)

Synchronized blocks were using different locks.
Replaced by a safer version: marking intern method synchronized.
This means synchronizing more than necessary but it had no
mesurable perfomance impact.

Bug 11744785

Change-Id: I35e691232cd6971d13735be9b72969739ef71e09
(cherry picked from commit 5c4883e7)

It should already have been the case, but it wasn't and it crashed.

Change-Id: Icb461d6e43ef2d113f59038d8c2327f3cfaff86c
(cherry picked from commit cf069ca5)

Bug: 11045348

Change-Id: Ib13f96563f3b7ead17620e409735a0e9abb2a109

Bug: 11045348
Change-Id: I287703eaa964dcbd2e8d5a6c7db7ac378e19e2c1

Bug 10907247

Change-Id: I0665cce648dd892db4cc59e82c325b376dad97dd

Bug 10907247

Change-Id: I0665cce648dd892db4cc59e82c325b376dad97dd

Bug: 10899202.
Change-Id: I6247f895a11b41a70d7f43562d868dd91e8b1a91

Macs don't have off64_t and instead use loff_t. This is taken care of in
the Compat.h header from system/core.  Include that.

Bug: 10424836
Change-Id: Ic892e1edfecab502159209d463394864370c5537

Copy new behavior of ZipFileRO to dalvik's ZipArchive.cpp

Bug: 10424836
Change-Id: Ic0ce11281f7318abda9d3c8062a095186448d2a9

Apps are no longer required to have sdcard_r or sdcard_rw to access
some parts of external storage.  Relax the mount parameters to allow
everyone to traverse.

Bug: 10577808
Change-Id: I620c8f02f08b149437dfa5705e05a27b6994d2db

Bug: 10134551
Change-Id: I9df8f7a2541e0c9bd2f9294a202fef2afce815f8

The OBB storage area is now presented on a per-user basis by the FUSE
daemon, so we no longer need to create bind mounts.  (In fact, we
must use the FUSE daemon since it now correctly synthesizes UID
ownership on a per-user basis.)

Bug: 10506809
Change-Id: I2d6c5643d7c89d7a15ceb28af871e44618ad33ad

Bug: 10649815

(cherry picked from commit 89ee8af8)

Change-Id: If28fab8c87222a9d182398a4deca167a8ca5c677

This allows traceview to tell whether tracing is active, and whether
it is sampling or method tracing.

Bug: 9968521

Change-Id: I72100d1536ea3168998110ec1cfa5a183b55a67c
(cherry picked from commit 9d3a0a2e253aecd07c4a053c19cf9b0ccaa2db49)

(cherry picked from commit 86bb14a3)

Change-Id: Ifece45f6a6644c20b6681a58dd3eae3534ad47e4

Instead of method instrumentation, allow traceview to periodically
sample stack traces of threads. Disabled by default until we add
gui support for this new mode.

(cherry picked from commit 676f8a52)

Change-Id: Ia5d0d57012305a5830d042bcb903a429432b035b

This restores the old message, replacing "ids" by "references" and
advertising for multidex otpions.

Updates dalvik/tests/089-many-methods accordingly.

(cherry picked from commit d352de04)

Change-Id: I5934bc9fb7812b66c0e38935946cacdf1c35a402

(cherry picked from commit 39bcbf75)

Change-Id: Iaddb2020b75f2bbba2e2ebda3a4465a890e361ca

- Classes in archives were added in all dexes.
- Adding resources could be done twice in non "strictNameCheck" mode
(when using BestEffortMainDexListFilter) if there was 2 secondary
dex.

(cherry picked from commit ad27e196)

Change-Id: I5cbc1b273518e868fdd3d642d50488ddcbec2af6

Add --multi-dex options to dx command line to allow the generation of
several dex files when method index limit is about to be reached.
Also add
  --main-dex-list allowing to force some classes in the main dex.
  --minimal-main-dex to keep in main edx only classes specified by
      main-dex-list.
  --set-max-idx-number to set an arbitrary idx limit for the
      splitting.

(cherry picked from commit c7daf656)

Change-Id: I2b42272be91484a75783eb94cd30581159948975

Fixed issue where we had inverted logic when checking for
pending exceptions.

Changed the logic to not update the native allocation watermarks
after concurrent GC since the finalizers may not have been run at
this point.

Copied over the test from ART for ensuring that it is working.

Bug: 10444866

Change-Id: Ie26170f9c4703f11e993233cdc9c5f46396fca84

Normally AndroidRuntime.cpp patches over this lapse, but that means
that tests run by vogar or any other kind of "bare" VM doesn't handle
SIGPIPE correctly. I'll remove the workaround from AndroidRuntime.cpp.

Bug: 10211309
Change-Id: I51380486867af011d58ddbefbdef8d120f736ad9