Replaced spring framework with robospring.
Changed ethereumj-core build to android-library.
Disabled BlockStoreImpl until we make it android compatible( use InMemoryBlockstore ).
Disabled antlr4 dependencies.
Commented some dependencies(needs testing before enabling).
Excluded dom4j and org.jboss.spec.javax.transaction dependencies (problems with android).
Change to netty 4.0.18.Final (this seems to work with android; latest 4.1 and 5.0 presented some problems; needs additional testing).
The sources affected by this commit have drifted in style since earlier
polishing commits such as f34d1f4 and a155518. The shared IDEA style
configuration file can help avoid this drift in the future. Simply run
Code->Reformat Code at the project level prior to committing. Note that
this code style configuration is not perfect, and not every change it
makes should be checked in. The ideal is to simply always write code that
conforms to the style conventions, but the automated settings are there
as a helping hand when necessary.
The churn caused by drifting style and the commits that fix it damage the
utility of git over time. Reviewing changes, analyzing history, surgically
reverting commits, and many other use cases with git simply do not work as
well when styles get mixed and fixed. Let's work toward a common style.
This is in order to accommodate GitHubStateTest#stSystemOperationsTest,
which tests Ethereum's rules around maximum call stack depth, and thus
requires increasing Java's own defaults.
If this option is not set, this test results in a
StackOverflowException. Note that the setting is applied to Gradle's
`test` closure as opposed to a the global `JAVA_OPTS` environment
variable, because Gradle spawns a new process for executing tests.
There is a known issue with Gradle's new `maven-publish` plugin that
causes all dependencies in the published pom to runtime-scoped [1]. This
causes conflicts with our use of Spring's `propdeps` Gradle plugin that
introduces optional and provided scopes.
This commit drops down to using Gradle's older, orginal `maven` plugin,
and JFrog's old-style `artifactory-publish` plugin in order to preserve
propdeps support.
This means removing JFrog's `bintray` plugin, as it appears to depend on
the `maven-publish` plugin. For the moment, ethereumj only really needs
to publish snapshots, and this is still supported. This is a stopgap
measure; it will be necessary to deal with getting uploads to
Bintray/JCenter working again once we're ready for a non-snapshot
release, but this can be dealt with later.
[1]: http://forums.gradle.org/gradle/topics/maven_publish_plugin_generated_pom_making_dependency_scope_runtime
Use Spring's `propdeps` plugin to introduce optional/provided scopes for
gradle dependencies. Mark libraries such as slf4j-log4j12 as optional.
These libraries are required to run a local ethereumj node, but they are
not required in order to compile against ethereumj in a downstream
appliaction. These libraries easily can cause conflicts, for example if
another slf4j-* library (such as slf4j-logback) is on the classpath.
Prior to this commit, jacoco coverage report XML files were not actually
being written to disk, meaning that the Gradle coveralls plugin had
nothing to publish to https://coveralls.io/r/ethereum/ethereumj. HTML
reports are now being written as well, for convenient local browsing.
The Travis CI build now publishes ethereumj-core artifacts (classes,
sources and javadoc jars) on every push to master. See updated
ethereumj-core/README.md for details. Note that the `gradle release`
task has been renamed to `gradle publish`.
This partially reverts commit c95f5ea, restoring use of Java 8 language
features and APIs to ethereumj-studio, while leaving etherumj-core
pinned to the Java 7 language level for Dalvik compatibility reasons.
For further details, see the prior commit and
https://github.com/ethereum/ethereumj/commit/c95f5ea#commitcomment-9119435.
As part of pull request #179, commits 0d922e1, 003249c and d099100
introduced use of Java 8 language features and APIs. Commit c95f5ea
manually reverted these changes (see #184 as to why), but in the process
re-introduced formatting issues originally cleaned up by other commits
in #179. This change fixes those formatting issues.
- Replace leading tabs with spaces (for reasons detailed in 0827fb5)
- Add space before opening curly brace
- Optimize imports using shared .idea/codeStyleSettings.xml (note
especially expansion of wildcard imports. See rationale in 780393d)
- Do not align assignments on equals sign
- Remove unnecessary additional newlines
- Remove braces from single-line loops and conditionals
- Replace leading tabs with spaces
- Add space before opening curly brace
- Optimize imports using shared .idea/codeStyleSettings.xml
- Do not align assignments on equals sign
- Remove unnecessary extra newlines
Running `gradle build` within ethereumj-core will now produce -sources
and -javadoc jars in `build/libs`, alongside the already existing
classes jar. Run `gradle install` to install these artifacts to the
local `~/.m2` repository.
This change ensures that all Javadoc syntax, reference, html errors, etc
are reported at the commandline during the gradle build, but warnings
for missing Javadoc elements (e.g. no @return tag, @param tags, etc) are
suppressed. `man javadoc` for details.
See updated ethereumj-studio/README.md for details. Note that this
commit removes the `build-post-package.xml` ant file (for now) to
simplify the Gradle arrangement. This functionality can be re-introduced
later and more simply using Gradle's native DSL. In the meantime, both
the -core and -studio projects can be run using `gradle run` or run as
executable jars by running `gradle shadowJar` followed by the usual
`java -jar` execution of the resulting `-all` jar file found in each
project's respective `build/libs` directory.
- Convert ethereumj-studio build from Maven to Gradle as well
- Extract common build elements to top-level build.gradle
- Update top-level README accordingly
Apply Gradle's 'maven' plugin in order to support `./gradlew install`
task, similar to `mvn install` for building and copying artifacts to the
users local ~/.m2 repository.
After porting the build from Maven to Gradle (see prior commits), the
org.ethereum.serpent.ParserGenerator class is no longer necessary as the
configuration of the antlr compiler that happened in this class now
happens directly in build.gradle. This class is thus removed in this
commit.
The plexus FileUtils class introduced by the antlr4 Maven plugin had
also been used in several locations in the codebase, probably just for
reasons of convenience. These uses have been replaced with Spring's
FileSystemUtils in order to allow for dropping the dependency on the
antlr4 Maven plugin entirely.
ethereumj-core depends on elements specific to the JUnit 4.11 API, but (for
reasons that can only be assumed dubious) json-simple:1.1.1 transitively
includes JUnit 4.10. At the command line, Gradle handles this cleanly,
and gives ethereumj-core's direct dependence on 4.11 the precedence it
deserves. However, IDEA gets confused and gives 4.10 precedence. This
change explicitly excludes the transitive dependency from
json-simple->junit:4.10, eliminating the issue completely.
Previously, sources generated from the serpent antlr grammar in
ethereumj-core/src/main/antlr4 were output into the build/generated-src
directory by default. This worked from the command line without issue,
but caused issues within IDEA, because the entirety of the 'build'
directory is excluded from compilation by default when working with a
Gradle-based project in IDEA.
This change simplifies the arrangement by writing these antlr-generated
sources to the src/gen/java. A .gitignore file has been added to ensure
that transient source files don't accidentally get checked in. A readme
file has been added to explain the situation to the uninitiated.
Adrian Tiberius
Roman Mandeleil
Devrandom
Faiz Khan
Chris Beams