Previously, if an implementation jar (Host) was not available to the
build it would panic when trying to generate the JSON representation of
the CLC. That prevents builds with missing implementation jars from
working even if those jars are never actually built.
This change defers the build failure until it is actually built.
Test: TARGET_PRODUCT=armv8 TARGET_BUILD_VARIANT=eng ./art/test/testrunner/run_build_test_target.py -j80 art-no-prebuild
- run above in master-art before and after this change.
Bug: 202366925
Change-Id: I60a78a8bf6c13b83a9dceb5c43019a9e21f0b637
These properties specify libraries that cannot be implicitly inferred by
Soong. If these properties are added to Android.bp, this can only be for
the reason that there is a <uses-library> tag in the manifest which is
unknown to the build system. Adding them to the manifest_fixer doesn't
make sense: if they are not in the manifest, they should be removed from
Android.bp as well.
Bug: 132357300
Test: $ lunch aosp_cf_x86_64_phone-userdebug && m && launch_cvd
$ adb wait-for-device && adb root && adb logcat \
| grep -E 'ClassLoaderContext [a-z ]+ mismatch'
# empty grep output, no errors
Change-Id: Ic6eb5268a954ef3be7f06a181ec72af99000c547
Previously manifest_fixer used a naive way to distiniguish optional libs
from required ones: it checked if a library is on the list of optional
compatibility libraries. This works for compatibility libs, but not for
other libs.
Now we properly track optionality through all stages of the build,
starting with the addition of the library as a dependency (here's where
the `uses_libs`/`optional_uses_libs` distinction kicks in), store it in
dependency tag and propagate to class loader context, and from there to
the manifest_fixer.
The tests have been updated accordingly.
Bug: 196377222
Test: lunch bertha_x86_64-userdebug && m droid dist cts mts
Change-Id: I3631ce59ebe47116ce7a9b3d33a86f636846ef0f
A Java module may depend on a stub library. In that case an additional
dependency on the implementation library is created, and it is used to
add the implementation library to class loader context. We should not
attempt to add the stubs library as well (previously the attempt to add
it happend after the implemention was added to CLC, to the attempt was
unsuccessful).
Raise an error if someone tries to add the same library with different
build/instal paths.
Also, rename local variable `implicitSdkLib` to `sdkLib` to better
reflect its meaning.
Bug: 193425964
Test: $ lunch aosp_cf_x86_64_phone-userdebug && m && launch_cvd
$ adb wait-for-device && \
adb root && \
adb logcat | \
grep -E 'ClassLoaderContext [a-z ]+ mismatch' -C1
# empty output, no errors
Change-Id: I01c1bdd23f9d118d891d0b806e7e3b4d78896a34
This is needed when some source libraries get replaced with prebuilt
variants: CLC should contain one entry for a library (the prebuilt one
or the source one, whichever of them is preferred). Because the prebuilt
module name starts with "prebuilt_" prefix, previously Soong considered
such libraries as two different ones libraries and added both to CLC.
Bug: 193425964
Test: mark "prebuilt_android.net.ipsec.ike" as preferred and `m nothing`
(before this CL it would fail the build, now it builds fine).
Test: $ lunch aosp_cf_x86_64_phone-userdebug && m && launch_cvd
$ adb wait-for-device && \
adb root && \
adb logcat | \
grep -E 'ClassLoaderContext [a-z ]+ mismatch' -C 1
# empty output, no errors
Change-Id: Icc42533d9915060d7fffda12aa93b9d18dc4f83d
toJsonClassLoaderContextRec's size is twice as large than expected. And
the initial values is filled with null value.
Bug: 158843648
Test: m nothing(testcase is added)
Change-Id: I24c23269bd16baa18481f34b85c543d41f26d0e0
Library order is important because it is used to construct class loader
context, which is then written into OAT/ODEX files and chacked against
class loader context constructed by PackageManager on the device. If the
orders are different, dexpreopted code is rejected.
Soong avoids using Go maps for class loader context representation
precisely for that reason. However, for the modules defined in makefiles
dexpreopt configs were serialized to JSON and unmarshaled to Go maps,
which resulted in wrong order of libraries. This CL changes Go
representation of class loader contexts imported from JSON and makes the
order stable.
Bug: 132357300
Test: lunch cf_x86_64_phone-userdebug && m && launch_cvd \
adb wait-for-device && adb root && adb logcat \
| grep -E 'ClassLoaderContext [a-z ]+ mismatch'
# empty grep output, no errors
Change-Id: I15f51617f9573c0bbcb324cf2592daf719cad586
This is needed for Java libraries that are <uses-library> dependencies
of Java libraries and apps defined as Make modules. Each dexpreopted
module in Make generates a dexpreopt.config file, which incorporates
information from its dependencies' dexpreopt.config files. For
dependencies that are Make modules their dexpreopt.config files are
generated by Make, and for Soong modules they are generated by Soong.
Since Soong doesn't know which libraries are used by Make, it generates
build rules for a superset of the necessary libraries.
Bug: 132357300
Test: lunch aosp_cf_x86_phone-userdebug && m
Change-Id: I325b1037658736ee3c02450b08c00eca1a175962
Previously Make class loader contexts used a simplified one-level map.
Now they use the same tree representation as in Soong. The difference
between Make and Soong represenataions is insubstantial and caused only
by the language differences between JSON and Go.
Bug: 132357300
Test: lunch aosp_cf_x86_phone-userdebug && m
Change-Id: Ia273afb7f026dbe3b25d4327d55143657b026b98
The removed API has been unused since https://r.android.com/1533342
(except for unit tests).
Changes in the unit tests reflect the change of API in
https://r.android.com/1533342: early errors caused by unknown library
paths at CLC construction time have been replaced with late errors at
the time of CLC validation.
Bug: 132357300
Test: m nothing
Change-Id: I739c7c41b6f882b7e28cdd6acd05961d754d8687
Previously CLC construction was scattered across different module types
and dependency tags. This CL moves all logic to one function, which
handles all special cases. This will allow to simplify CLC API and
reduce the number of different ways in which CLC is constructed.
Previously some of the cases failed early (at the time when a library is
added to CLC) if the build/install paths were unknown. Other cases did
not fail early, but were validated later before CLC was used. Late
failures are necessary because some of the libraries with unknown paths
still have to be processed by manifest_fixer (which doesn't need library
paths), but they do not use dexpreopt (which needs library paths). This
CL removes the early failures (all paths are still validated later).
The CLC tests do not fail because they use a private method that toggles
the "strict" flag (that enforces early/late failure mode) manually in
the method call.
The CL also makes a functional change in the way CLC is constructed for
component libraries that have an OptionalImplicitSdkLibrary(), or
libraries that are disguised as SDK libraries via `provides_uses_lib`.
Previously such a component/disguised library X was added to its own CLC
as a sibling element of X's own <uses-library> dependencies, which
created incorrect CLC structure. Now this is handled by addCLCFromDep,
when X is processed as dependency and added as a top-level CLC element
with its sub-CLC properly nested under it.
Bug: 132357300
Test: lunch aosp_cf_x86_phone-userdebug && m
Change-Id: I6a512209b87b81d785875f10f76b21c81b2ed579
This patch reworks the approach introduced in
https://r.android.com/1450819. That patch based the decision which
libraries should be added to the manifest <uses-library> tags by the
manifest_fixer on the "shared" status of the library.
That approach is incorrect for two reasons:
- It doesn't make sense to have a non-shared library in class loader
context ("shared" libraries are those specified in
frameworks/base/data/etc/platform.xml, and they are the only ones
that PackageManager knows about).
- It doesn't make sense to add anything but the top-level of the
class loader context tree to the manifest, because this part of the
tree is flattened to a sequence, and PackageManager cannot restore
it to the previous tree shape (there is an information loss).
This patch removes the "shared" bit of information from class loader
context elements and assumes that all libraries that end up in class
loader context are shared. Consequently, only the top-level libraries
should be passed to manifest_fixer.
Test: lunch aosp_cf_x86_phone-userdebug && m
Bug: 132357300
Bug: 168686456
Change-Id: I902690f0f38f1047fa79cf6ccbe956077eceaab0
When adding a subcontext in a class loader context tree, there are two
possible cases: 1) the root of the subcontext is itself a <uses-library>
and should be present as a node in the tree, or 2) the root is not a
<uses-library>, but some of its dependencies are -- in that case they
should be disconnected from the root, and the resulting forrest should
be added at the top-level.
Example:
1) C is a <uses-library>:
A
├── B
└── C
├── D
└── E
└── F
2) C is not a <uses-library>:
A
├── B
├── D
└── E
└── F
Before the patch subcontexts for transitive dependencies were added
before the subcontext for the direct dependency (even if it was a
<uses-library>, resulting in case-2 hierarchy when case-1 should have
been used. Previosuly this didn't matter because class loader context
was a flat set of libraries, but now it matters because class loader
context is a tree.
This patch changes the order in which libraries are added, so that
direct dependencies are added before transitive ones. The context adding
method now accepts an "implicit root" parameter, so that when adding
transitive dependencies it can check if the corresponding direct
dependency is a <uses-library> and already present in the context.
Partially constructed class loader context is now propagated top-down
into aapt.buildActions, so that the method can use existing part of the
context to decide where the missing part should be connected.
Test: lunch aosp_cf_x86_phone-userdebug && m
Bug: 132357300
Change-Id: I649aff9e27494306885a4f4fc90226c399636b57
Nested conditional context doesn't make sense because conditional
context is only needed for compatibility libraries, and those are only
used by apps, and apps are always at the root of class loader context
tree. Therefore, trying to add conditional nested context can only mean
an error somewhere on the way.
Test: lunch aosp_cf_x86_phone-userdebug && m
Bug: 132357300
Change-Id: Iadeaaf89fdb11cb23b107c6fb074b1bc765256be
The old representation consisted of a list of libraries (UsesLibraries),
a list of optional libraries (OptionalUsesLibraries) and a mapping from
library name to its build/install paths (LibraryPaths). The separation
into lists and map was necessary because of special handling of
compatibility libraries, which is now unified with normal libraries.
The new representation is a mapping from target SDK version to a tree
structure ClassLoaderContext. Each node of the tree represents a library
and contains library name, build/install paths and a slice of
subcontexts for dependencies. The same library may occur in the tree
multiple times in case it is a dependency of multiple libraries. The
order in which libraries are added matters (the resulting tree shape may
be different).
Test results have to be updated, as the resulting <uses-library> list is
reodered (previously it was a sorted list of map keys, and now it is
formed by a depth-first preorder traversal of the class loader tree).
Test: lunch aosp_cf_x86_phone-userdebug && m
Bug: 132357300
Bug: 168686456
Change-Id: I11be8cd2967f004fd58753d7c5fb99fed179cd63
This dependency is incorrect and shouldn't exist.
It has been removed in https://r.android.com/1467918
Test: m nothing
Bug: 170710203
Bug: 132357300
Change-Id: If3036437e138b552436f135425e6bd15be043678