* Remove UnsafeReflectionAccessor
Revert #1218
Usage of sun.misc.Unsafe to change internal AccessibleObject.override field
to suppress JPMS warnings goes against the intentions of the JPMS and does not
work anymore in newer versions, see #1540.
Therefore remove it and instead create a descriptive exception when making a
member accessible fails. If necessary users can also still use `java` command
line flags to open external modules.
* Fix failing to serialize Collection or Map with inaccessible constructor
Also remove tests which rely on Java implementation details.
* Don't keep reference to access exception of ConstructorConstructor
This also avoids a confusing stack trace, since the previously caught
exception might have had a complete unrelated stack trace.
* Remove Maven toolchain requirement
* Address review feedback
* Add back test for Security Manager
Adversaries might be able to forge data which can be abused for DoS attacks.
These classes are already writing a replacement JDK object during serialization
for a long time, so this change should not cause any issues.
Previously ISO8601UtilsTest.testDateFormatString() would fail on systems
where the time zone is UTC+X because getTime() returned "2018-06-24" for them.
Additionally the tests which previously changed the system locale and time
zone have been rewritten to create a UTC calendar instead. Setting locale
seems to not be necessary because ISO8601Utils.parse(...) does not do that
either.
On some versions of Android (probably on some variants of the popular Samsung S4 phone), an older version of Gson is suspected to be bundled in, and gets picked up from the system classpath.
For those versions, the applications that include the latest Gson fail unexpectedly. This debug print will help confirm this issue.
* Fixed https://github.com/google/gson/issues/1310
Also renamed VersionUtils to more readable abstraction JavaVersion
Added support for debian naming convention
Using min supported version (6) as the default if JDK version can't be figured out
* Moved JavaVersion to an internal package
Added support for a magic field GSON_TYPE_ADAPTER in a class. This adapter is automatically invoked if present.
The field must be present in the class (not in any super-type), and must be strongly typed as TypeAdapter<T>.
The underlying problem is that the doubleCapacity function would drop the parent links when all nodes ended up on the same side in a doubling. This was caused by the fact that the AvlIterator was destructive on parent nodes, and we weren't putting them back together with the AvlBuilder. This removes an incorrect optimization and fixes the problem.
Also move LinkedHashTreeMap back into main from test.
The most interesting optimization is to replace ArrayDeque with a manual linked list that reuses the nodes 'parent' field. These optimizations save about 20%.
Compared to LinkedTreeMap, this is slower for small (size=5) maps: 124% slower to get() and 33% slower to create and populate. It's a win for large (size=500) maps: 46% faster to get() but 8% slower to create and populate. And it's a big win for very large (size=50,000) maps: 81% faster to get() and 46% faster to create and populate.
http://microbenchmarks.appspot.com/run/limpbizkit@gmail.com/com.google.common.collect.MapBenchmark
I'm going to follow this up with some simple optimizations: caching local fields and simplifying access. That should narrow the performance gap.
Not yet adopted in our code.
Known critical bugs:
- throws ClassCastException when get() is called with a non-comparable key
- throws NullPointerException on get(null)