Rtti was already handled by bazel macros, so this change simply
propagates the rtti bit to these macros.
Test: Run bp2build, build //external/libcxxabi:all
Change-Id: I63296db2db868202874c8bd0b1de6310f7bb85c7
This serves to not link parts of soong_ui (and eventually soong_build)
into a separate, weird binary. This is in turn good because they contain
any number of global variables and no one really thought about what
happens when two instances are executing at the same time in the same
address space.
This comes with a slight performance hit: 5 aosp_* projects build 152
seconds instead of 146. I suppose this is a price worth paying for a
clean design?
Test: presubmits.
Change-Id: I5623dcab2290f0fc392dd2ede597b9794a3d2a4e
This is in preparation to r.android.com/1740313 where setting correct
variables would be enforced (i.e. apex and non-apex boot jars must be in
config.ApexBootJars and config.BootJars correspondingly).
Bug: 191369843
Test: m nothing
Change-Id: Ic86680c1f7af53d229083b2cc58beb3ceccb4b6a
Previously, the signature-patterns.csv file included a lot of
implementation details, e.g. the signatures of dex members or inner
classes that are not part of any API, including the hidden API.
This change will remove all member signatures and inner class names
from the file and replace them with just the outermost qualified class
name.
That will still leave some implementation details, e.g. the names of
implementation only classes and packages.
Bug: 194063708
Test: atest --host verify_overlaps_test signature_patterns_test
m out/soong/hiddenapi/hiddenapi-flags.csv
- manually change files to cause difference in flags to check
that it detects the differences.
Change-Id: I9de6a2a6129e875e19f7ded5fae578cbdb584660
This CL removes the roots section from the rust-projects.json file as it
is no longer required by the rust-analyzer tool.
Bug: 162055556
Test: prebuilts/rust/tests/rust-analyzer-integration-test.sh
Change-Id: I85c3277632c88394bc52e709ecce83f6aa1f2e0e
Previously, a simple map from dex signature to the flags was used to
store the monolithic hidden API flags. This change replaces that with a
trie that uses packages and class names extracted from the signature
to map from the signature to the flags.
The signature is broken down into pieces consisting of package names,
class names and the remaining member signature. They are used in turn
to navigate through nodes in the trie to a Leaf that contains the flags
associated with the signature.
The flags can be retrieved from the trie in a number of ways:
* Using the complete signature to retrieve the flags for a single
member.
* Using a class name to retrieve the flags for all the members in a
class and any nested classes.
* Using a package wildcard to retrieve the flags for all the members in
the classes in that package.
* Using a recursive package wildcard to retrieve the flags for all the
members in the classes in that package and nested packages.
This will allow a bootclasspath_fragment to select a subset of the
monolithic flags to compare against itself without specifying the
signatures of every member of that set.
Before creating the trie an attempt was made to compute the subset by
iterating over all the signatures in the monolithic flags and matching
against a regular expression created from the patterns but that was too
slow. It took minutes to run whereas using the simple map or the trie
only takes a few seconds.
Bug: 194063708
Test: atest --host verify_overlaps_test
Change-Id: I36f5e319d3e7d62dd34305de1eec990a93cb3a89
- If no system_shared_libs is specified, bp2build writes no attribute
value. In this case, the bazel library macros determine the correct
default behavior.
- If any system_shared_libs is specified for any variant, then bp2build
writes the value verbatim. This includes if an empty list is specified,
as this should override defaulting behavior.
Note this defaulting behavior is incomplete and will be incorrect in
corner cases. For example, if, in an Android.bp, system_shared_libs is
specified for os.linux_bionic but not for os.android, then the bazel
default for os.android will be incorrect. However, there are no current
modules in AOSP which fit this case.
As a related fix, supports static struct for cc_library_static.
Also, removes some elements from the bp2build denylist.
Test: mixed_droid CI
Change-Id: Iee5feeaaf05e8e7209c7a90c913173832ad7bf91
Revert submission 1743633-switch-to-clang-r428724
Reason for revert: fix build
Reverted Changes:
I824f3bc1c:Update bazel's clang version to clang-r428724
I6c4c26267:Adjust test for compiler update.
I94ab1f2dc:Switch to clang r428724 13.0.1.
Change-Id: I846e904b6e868d4e2be1bd8260ed59fc5464ef68
Bug: 194063708
Test: atest --host verify_overlaps_test signature_patterns_test
m out/soong/hiddenapi/hiddenapi-flags.csv
- manually change files to cause difference in flags to check
that it detects the differences.
Change-Id: Ic819def9b0c41e3b22fb98582cd39f6d27080a32
Previously, the signatures used to select the subset of the monolithic
flags were simply the signatures read from the modular flags file. This
change moves the creation of the signature list into a separate script
that outputs the signatures to a file and then passes the path through
Soong from the bootclasspath_fragment modules that create it to the
platform_bootclasspath module that uses it to compare the modular
flags against the monolithic flags.
Currently, the signatures are the full signatures but follow up changes
will replace them with patterns (hence the name) that avoids having to
include implementation details in the hidden API flags that are output
as part of a bootclasspath_fragment's snapshot.
This change moves the stub flags related code next to the all flags
related code as they are treated in a similar way.
Bug: 194063708
Test: atest --host verify_overlaps_test signature_patterns_test
m out/soong/hiddenapi/hiddenapi-flags.csv
- manually change files to cause difference in flags to check
that it detects the differences.
Change-Id: I2855bf6d05c91b8a09591664185750361c7e644f
R8 will parse proguard files out of resources in injars. Use the jar
with resources instead of the classes jar so that R8 can see the
proguard files.
R8 fails when an input jar contains dex files, so filter out dex files
from the input jar before passing it to r8.
This relands Ibb870ee9c70470336f542a3b7542dab86716dbf8 with a fix to
move the temporary input jar outside the output directory. Putting it
in the output directory interacted poorly with RBE in local comparison
mode, which incorrectly deleted the output directory.
Bug: 195558228
Test: m checkbuild
Change-Id: Id4b091c3bd72d4c2904883e8793ec4b9499e4b78
This reverts commit 29c294b2cf.
The deadlock in R8 was fixed, and the timeout causes problems when
RBE is enabled and the R8 action gets stuck in a queue for 30
minutes.
Test: checkbuild
Change-Id: Ia0c3826aea3288e9db3bdbd4955e1e16c0a38fe3
Print more messages in run_with_timeout to make it easier to
determine which log statements came before the timeout and the
on_timeout command.
Test: run_with_timeout_test.go
Change-Id: I92ad663529fa03d91185ba382654a73f0eba1ab7
Previously, the compare_signature_flags compared the full set of
monolithic flags against the modular flags and just ignored any
monolithic flags for a signature that was not present in the modular
flags. It essentially performed two tasks, identifying the subset of
the monolithic flags against which the modular flags should be compared
and then doing the actual comparison.
This change separates those two tasks, by first creating a dict that is
the subset of the monolithic flags against which the modular flags will
be compared and then do the comparison.
At the moment the subset is simply the set of monolithic flags
associated with the signatures that are present in the modular flags.
Follow up changes will provide other ways to select the subset.
Bug: 194063708
Test: atest --host verify_overlaps_test
m out/soong/hiddenapi/hiddenapi-flags.csv
- manually change files to cause difference in flags to check
that it detects the differences.
Change-Id: Icb547f071b814dd955f31255b0d35597f7ba4aa1
Switches symbols to use monolithic... and modular... to try and clarify
the purpose of those symbols.
Bug: 194063708
Test: atest --host verify_overlaps_test
Change-Id: Ifc646edebe3d43b51dc3528fa883718399cb5254
Refactor verify_overlaps to make it testable and add tests for the
comparison. It does make one significant change in behavior which is to
read each of the files produced by a bootclasspath_fragment into a dict
before comparison, rather than reading and comparing them a row at a
time. That allows it to reuse the code to read a CSV into a dict.
Bug: 194063708
Test: atest --host verify_overlaps_test
m out/soong/hiddenapi/hiddenapi-flags.csv
- manually change files to cause difference in flags to check
that it detects the differences.
Change-Id: Ib70ac87fe089fc25e3bef18f367d4939bfc0cb8d
R8 will parse proguard files out of resources in injars. Use the jar
with resources instead of the classes jar so that R8 can see the
proguard files.
R8 fails when an input jar contains dex files, so filter out dex files
from the input jar before passing it to r8.
Bug: 195558228
Test: m checkbuild
Change-Id: Ibb870ee9c70470336f542a3b7542dab86716dbf8
Calling this function for dependencies that are not android.Module
results in a hard error rather than a soft skip. Update the doc to
reflect this
Bug: 195369546
Test: TH
Change-Id: I18a0938c6536cf20f239cf98c601722aaecd51e6
This special-casing is actively unhelpful nowadays, as it breaks
callstack unwinding and symbolisation by tools that work using the
on-device debug info (e.g. perfetto's profilers). It's also highly
surprising that eng builds have less debug info than userdebug/user.
As far as I can tell, the history behind this case is (with some
internal-only references as the context is in the review comments): the
minidebuginfo was originally rolled out in device-specific makefiles
(e.g. ag/2433691). Eng builds were excluded, presumably due to time/size
concerns for rebuild+resync workflow, as well as the assumption that
engineers using eng builds already have the necessary debug info
off-device. Later, as some more makefile variables were introduced, the
conversation in ag/990059 shows that the overheads were no longer a
concern, and the last patchset generalised the option to include eng
builds. However at a later point, during device-specific makefile
cleanup, the eng-exclusion got restored: ag/2455645. This then ended up
in the makefile->soong transition, leaving us with the current special
casing that is undesireable.
Checking things with an internal master build on my workstation:
dexpreopting system services is 10s regardless of this minidebuginfo
flag, and the resulting odex size is 37MB vs 38MB. Overall, I only see 4
dexpreopt invocations as part of a local build, so the effect of this
patch should be negligible.
Bug: 194835631
Change-Id: I80de75a84259c056cdae0696fe611afe7e0162f5
