org.apache.ofbiz.base.start.
(OFBIZ-9573)
For r1815119 I wrote:
No functional change, while reviewing r1811413
I noticed we can use a try-with-ress here
Fixes a possible not closed
As reported by Michael on dev ML:
if I see it right, you have not implemented a try-with-resources block but
simply moved the stream declaration inside the try block and removed the
finally block. Please check.
Michael was right, here is the fix
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/ofbiz-framework/trunk@1818514 13f79535-47bb-0310-9956-ffa450edef68
org.apache.ofbiz.base.start.
(OFBIZ-9573)
I slighly modified the patch because of a conflicting change.
Thanks Dennis Balkir for reporting and providing the patch.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/ofbiz-framework/trunk@1811413 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-9441)
A major refactoring patch that applies the following changes:
- Fully refactor the EntityDataLoadContainer
- Fully refactor the TestRunContainer
- In both above mentioned containers the logic is broken up into smaller
methods to enhance code readability. It also changes the signature of
the method calls to rely on StartupCommand instead of String[] arrays
- Remove most old documentation and commented out code where applicable
- Delete the StratupCommandToArgsAdapter as it is no longer needed
- Create a new feature in which OFBiz by default fails if any data file
fails to load. This can be overridden by passing a flag called
"continue-on-failure" e.g. ./gradlew "ofbiz --load-data continue-on-failure"
- Add a new property to --test called "loglevel" (old code but working properly)
- Add many new properties to the --load-data command including repair-columns,
try-inserts, maintain-txs, etc ... These properties are old, but somehow
did not make it into the command line arguments. This code is now exposed
through the command line to set a direction in the future of whether to
keep or delete these extra commands
- Update the documentation output of ./gradlew "ofbiz --help" to incorporate
the new mentioned properties
- Refactor README.md to incorporate the changes to the gradle commands for both
"-load-data" and "-test" properties. Also remove the gradle GUI
documentation as it is now deprecated.
- Also refactor README.md in other locations to cleanup and make it more
consistent. This icludes moving long notes into new sections and reducing
the verbosity of the security header. Furthermore, created a new header
called Miscellaneous documentation to house the newly created sections
- lots of small changes to remove EOL white space
Thanks: Jacques Le Roux for reviewing and testing
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/ofbiz-framework/trunk@1805947 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-9571)
No functional change.
We read at http://docs.oracle.com/javase/7/docs/api/java/math/BigDecimal.html :
<<Using the integer fields in this class (such as ROUND_HALF_UP) to represent
rounding mode is largely obsolete; the enumeration values of the RoundingMode
enum, (such as RoundingMode.HALF_UP) should be used instead.>>
That's an easy global replacement. But it also needs to import the RoundingMode
class. It was 26 files so I decided to rather clean all the imports in all files
with the help of Eclipse.
This is a 1st step. I will need to closely look at 2 special class: UtilNumber
and Calculate.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/ofbiz-framework/trunk@1804953 13f79535-47bb-0310-9956-ffa450edef68
When reviewing code, Eclipse complains: "Javadoc: Invalid param tag name"
when a param tag name "contains" a colon at end
Weirdly it does not complain about @return. Go figure! But I'll apply the same!
This removes all cases found in trunk. Another way could be to add a space
between the tag and the colon. But that's not how it's recommended to write
http://www.oracle.com/technetwork/articles/java/index-137868.html
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/ofbiz-framework/trunk@1804637 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-9435)
Another round of refactoring in the start component with the following changes:
- Change the default adminKey to "NA" since it is only used in the "start"
context. Other contexts do not require an admin server (test, load-data)
- Simplify and cleanup the locale creation logic in the constructor and move
things out to the method with a return value
- Simplify the Timezone creation logic by moving it outside the constructor into
a small method
- Substantially simplify the determineOfbizPropertiesFileName(..) method by
reversing the order of condition evaluation (first, load-data, then test, then
anything else is start)
- Change default locale to be 'en' as the system crashes if not provided.
- Remove all comments from properties files except for start.properties and
refer to it from the other properties files
- Update the properties files accordingly
Thanks: Jacques Le Roux for review and suggestion to remove the comments from
properties files
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/ofbiz-framework/trunk@1800368 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-9285)
Provide log message after all containers/components are loaded.
This helps to inform new developer that OFBiz has started.
Thanks: Jacques and Taher for code review and mentoring.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/ofbiz-framework/trunk@1789533 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-9196)
Disabled the admin server from starting on contexts other than "start"
This commit has reference to the below discussion thread in which buildbot
is failing in between commits due to blocking the admin server thread.
The commit enables the admin server only for the "start" ofbiz server command.
For all other commands the port is set to value zero and hence the admin server
is never initialized
https://s.apache.org/xfxS
Thanks: Taher for the fix
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/ofbiz-framework/trunk@1787627 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-9155)
Symptom:
* services that are invoked with "dispatcher.runAsync" stay "pending" in the
job-list and never get executed
* scheduled jobs don't get executed
Steps to Reproduce
* Login to /webtools
* Schedule a Job to run in 5 mins, (for example service "clearAllEntityCaches")
* Wait 7mins
Expected:
* the scheduled service "clearAllEntityCaches" is not "pending" anymore but
should have status "finished"
Actual:
* scheduled service "clearAllEntityCaches" is still "pending"
Analysis:
The problem can be reproduced with the OFBiz 16.11 branxh as well as
with the OFBiz 16.11.01 zip file version and the trunk
JobPoller does get started on OFBiz start but does get stuck on line:
org.apache.ofbiz.service.job.JobPoller.java: 213-215
The reason is that the "Start.ServerState" never reaches "RUNNING"
Doing a full-text scan through the whole source code shows, that there is no
code at all that sets the ServerState to "RUNNING"
Comparing it to the 15.11-branch. There used to be a line:
serverState.compareAndSet(ServerState.STARTING, ServerState.RUNNING)
no such line exists in 16.11 anymore.
jleroux: I confirm the behaviour in R16 branch as well as in trunk.
Leonard's "patch" is a good solution and works for the R16 branch which is less
changed than the trunk when compared to R15.
The same applies to the trunk with the relative block of code which contains
if (serverState.get() == ServerState.STOPPING) {
The line Leonard propose to add was obviously missed during the recent
refactoring.
Thanks: Leonard Lin
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1779596 13f79535-47bb-0310-9956-ffa450edef68
Improvements remove in total roughly 70 lines of code and includes:
- moving all default values to the constructor to make it clear
- refactor getProperty to use the Optional monad
- remove FileInputStream from getPropertiesFile as it is not used
- properly indent conditions in determineOfbizPropertiesFileName
- simplify locale creation
- remove dependency on base (StringUtil)
- other minor renamings and formatting
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1775340 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-8337)
This is another major commit to the refactoring of the startup sequence in
OFBiz with the following highlighted changes:
- Delete the NativeLibClassLoader.java
- Delete the native classpath logic in Classpath.java
- Refactor the ComponentContainer and StartupControlPanel to operate without
the NativeLibClassLoader. This substantially simplifies the code
- Declare a URLClassLoader in ComponentContainer that is instantiated upon
building the classpath for all components. This makes the classloader start
in one shot
- Simplify the Config file and remove fields that are not used. Also
refactor some messy logic for loading the props file and other code
improvements
- Refactor all the .properties files for startup to have a consistent
structure that clearly documents all available properties and the
default value for each property. This change is also related to the
changes applied on Config.java
- Remove the declaration of the StartupLoader implementation class
from all startup .properties files because only one implementation exists
and it should be the default always.
- Refactor the loaders retrieval code (main, rmi, test, load-data) defined
in the startup .properties files
- Refactor some switch statements to comply with java coding standards
- Add the DTDs defined in base through Gradle because we removed the
NativeLibClassLoader and the classpath buildup logic in StartupControlPanel
Thanks: Jacopo Cappellato for reviewing the work
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1772879 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-8337)
The above mentioned revision introduced a regression in which starting
OFBiz in debug mode triggers classpath exceptions due to not finding
certain jars. Upon investigation we realized that the new logic
instantiates the loader class before the custom ClassLoader is
instantiated. The fix would take a bit more time, hence reverting now
to quickly resolve the issue and make the code base ready for the next
release. The revert essentially re-introduce a List<StartupLoader> data
set instead of a single StartupLoader.
This is a minimal revert (less than 400 lines) instead of a full revert
going above 2000 lines because we isolated only the parts that trigger the
regression mentioned above. However, we should ideally revisit this code
and clean it up again with careful attention to the class loading sequence.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1767866 13f79535-47bb-0310-9956-ffa450edef68
(OFBIZ-8337)
This is a major refactoring exercise to simplify the startup sequence in
OFBiz. There are no functional changes, however, both the logic and the
API signatures of the core classes in OFBiz are changed to help simplify
the startup sequence. The changes insofar are summarized as:
- Rename ContainerConfig.Container to ContainerConfig.Configuration, the old
nameing was very confusing as it confuses it with actual containers
- Change the signature of ContainerLoader and Container to accept
List<StartupCommand> instead of String[]. And also migrate the adapter for
converting these args into a standalone class called
StartupCommandsToArgsAdapter
- Remove the LockedBy annotation as it is not used
- Substantially simplify the ContainerLoader load and unload logic and delete
the start logic (merge it with load). The interface StartupLoader was updated
accordingly
- Remove old logic to load containers in hot-deploy components
- Remove printThreadDump output from ContainerLoader which used to exist for
debugging purposes and is no longer necessary
- Cleanup the AdminServer startup logic
- Change the StartupControlPanel logic to use only one StartupLoader instead
of a list. The only implementation is ContainerLoader
- Remove the synchronized blocks in StartupControlPanel due to having only one
loader
- Update the startup .properties files to reflect the changes in the startup
logic
- Update the TestRunContainer to use List<StartupCommand> in its logic
We expect more commits (specifically to the containers) to complete this
refactoring initiative
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1765127 13f79535-47bb-0310-9956-ffa450edef68
There were two tests in the start component that have a hidden regression which
can be repeated as follows:
- start ofbiz with a normal ./gradlew "ofbiz --start"
- Make some changes to the code
- stop ofbiz with ./gradlew "ofbiz --shutdown"
OFBiz will fail to stop because the unit tests will fail due to changed code
while OFBiz is still active.
The correct solution is to convert these tests to integration tests and handle the different
scenarios while the AdminServer is up.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1757360 13f79535-47bb-0310-9956-ffa450edef68
Removes the ofbizSecure and ofbizBackgroundSecure and makes all OFBiz server tasks
secure by default. Also, update the documentation of README.md and the StartupCommandUtil
to remove these entries from the documentation.
This has reference to a thread discussion here -> http://markmail.org/message/zbbun7zrzdrtxavv
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1756305 13f79535-47bb-0310-9956-ffa450edef68
Notes:
* Also removes empty framework/base/lib/scripting and useless applications\content\lib\uno\README (miss doc in wiki)
* When jpim-0.1.jar will be (hopefully) replaced we can drop lib ref in framework/base/ofbiz-component.xml
* I wonder if we should let the geronimo component alone with only 2 classes, could we not put that in entity or entityext?
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1755306 13f79535-47bb-0310-9956-ffa450edef68
Apply the following:
- remove the testlist command from options available in OFBiz server
- Delete the TestListContainer.java which implement the behavior
- Update the Config file to ignore properties related to that command
- Delete the properties file related to the command.
This has reference to the above JIRA and the below discussion
thread in which the community agreed to remove the testlist
server command from OFBiz. This command is an old and deprecated
command that used to create an ant script which iterates over all
test suites and execute them consecutively. It is both not-needed
and also deprecated given its dependence on Ant.
The discussion thread:
http://ofbiz.markmail.org/message/r64igaf6ihvpolaa
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1754507 13f79535-47bb-0310-9956-ffa450edef68
Most is done already. Test pass but testFopMacroLibrary but I don't see how it could be related to these changes, could be a quirk on Windows, BuilBot will tell us. Anyway, I will later double-check to be sure, also to see if I did not miss anything else.
I have voluntarily left 2 or 3 moot points
I was unable to remove possible POS references in <property key="FieldDescription.ProductStore.isImmediatelyFulfilled"> for certain language I'm unable to read.
I have to check if we should keep or not this code snippet
<!-- if inventory info is not available for all of the products, then don't do the issuance, ie there has to be at least SOME inventory info in the database to facilitate inventory-less POS terminals -->
<entity-count entity-name="InventoryItem" count-field="iiCount">
<condition-expr field-name="facilityId" operator="equals" from-field="orderHeader.originFacilityId"/>
</entity-count>
I simply commented out twice this snippet in InvoiceServices.java, because I wonder if it would not be useful when the webPos will be able to issue money change
// if (null == adj.get("amount")) { // JLR 17/4/7 : fix a bug coming from POS in case of use of a discount (on item(s) or sale, sale here) and a cash amount higher than total (hence issuing change)
// continue;
// }
While at it, since I modified OPTIONAL_LIBRARIES, I also changed these svn:properties: svn:eol-style = native & svn:mime-type = text/plain
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1754402 13f79535-47bb-0310-9956-ffa450edef68
The two tests were lumped together and if one throws an exception the other
will never throw it. So test coverage was not done properly. This commit
fixes this issue by separating the calls into two different tests.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1753348 13f79535-47bb-0310-9956-ffa450edef68
This is the first batch of unit tests for the Start component
which introduces 8 unit tests for parsing the start commands
and also some tests for the AdminClient.
At some point if the tests become numerous we might need to
break it down into multiple files. The tests are implemented
with standard JUnit tests. A minor correction was done to the
StartupCommandUtil class by converting the List of startup
commands to a Set to remove duplicates. This was discovered
thank to the unit tests being introduced.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1753274 13f79535-47bb-0310-9956-ffa450edef68
This is the first of two steps in the above mentioned JIRA. This
first step changes the java source directory for every component
from /src to /src/main/java. The next step will be adding unit
tests to every component in /src/test/java
Thank you to Deepak Dixit for tips on how to use the SVN commands
to make the appropriate file moving.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1752838 13f79535-47bb-0310-9956-ffa450edef68
This commit represents a pivotal change in the history of the
project in which the build system is being replaced from Apache
Ant to Gradle of which the discussion mail thread link is below.
Thanks to everyone who helped with testing and feedback
This commit delivers the following:
- add the gradle wrapper directory in gradle/* which holds
all items necessary for gradle to run
- add the build.gradle master script
- add the settings.gradle script which declares all the subprojects
- add documentation for how to use everything in a new file called
README.md (written in markdown)
The new script is capable of the functionality that used to exist
in Ant with some additional features including automation of eclipse
files generation and many other functions.
To complete this new feature, we need to apply the following _after_
this commit:
- change the buildbot to utilize gradle instead of ant
- delete all ant & maven scripts
- replace all local jars with remote ones from jcenter
Reference thread: http://ofbiz.markmail.org/message/vstt3wxuubmjgmqj?q=Important+Changes+to+Trunk+and+Use+of+Ant+%26+Gradle
Reference JIRA: https://issues.apache.org/jira/browse/OFBIZ-7534
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1751309 13f79535-47bb-0310-9956-ffa450edef68
I kept things as simple as possible. I only changed the following Ant targets
start
start-secure
start-batch
start-batch-secure
run-test
I also changed the startofbiz.bat file. The other start*.bat are for the POS and it's not expected to run on a server.
I put a mention when using "java -jar ofbiz.jar -u" (start)
I will add a mention in "Technical+Production+Setup+Guide" wiki page about using "java -jar ofbiz.jar"
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1748173 13f79535-47bb-0310-9956-ffa450edef68
A thank you is due to Jacques Le Roux, whom substantially improved
the refactoring of the start component with his QA work, and also
credit is due to Jacopo Capellato and Gil Portenseigne for their help
This commit is the last commit on OFBIZ-6783 and applies the following:
- convert the startup loaders from ArrayList to List. Given that
only two loaders existed in the system since the beginning of
OFBiz means that it is really unnecessary to use the
loaders.trimToSize() method. It's simply too much specificity for
no apparent performance value. this leads to changing many method
signatures from ArrayList to list across different classes
- More simplification of Config.java by simplifying timezone setting
logic
- remove the throws StartupException from main. This was confusing
confused me a lot and lead to a regression from my refactoring
in OFBIZ-7167 because some exceptions bubble up all the way to
main but then the system does not terminate because other
threads are running. It took me and Jacques a long while to
dig it out and figure the root cause.
- Introduce a function called fullyTerminateSystem(StartupException e).
This function replaces the exception bubble issue that is
described above. Consequently, the methods init, status and
shutdown no longer throw an exception but instead terminate
the system (which is the right thing to do).
- finetune adaptStartupCommandsToLoaderArgs(...) and other
places in StartupControlPanel
- separate the Classpath creation from NativeLibClassLoader
creation. This makes it cleaner and easier to refactor
in the future
- break the StartupControlPanel.start(...) method into
smaller more readable methods. It is easier now to understand
the startup sequence for ofbiz in english words (the method names).
It also allows for cleaner refactoring in the future.
- add JavaDoc to a lot of methods all over the start component
- rename AdminPortThread to AdminServer and create a new class called
AdminClient. Now everything relating to controlling OFBiz directly
happens through StartupControlPanel such as start, stop or init
for the server, and everything which happens indirectly through
the network (by an admin client) is done through the AdminClient
which talks to AdminServer
- more refactoring of AdminServer to simplify the code and break it
down into multiple methods and refactor some messy old code
Assuming this commit is stable, and after waiting for a sufficient time
I will close the JIRA and move on with the refactoring exercise.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1747288 13f79535-47bb-0310-9956-ffa450edef68
Thanks to an excellent investigation by Jacques Le Roux we realized
that if not all tests are successful (on java -jar ofbiz -jar --tests)
then the system hangs and does not print HTML reports.
After investigating this issue, I realized that the old logic before
refactoring used to call System.exit(99) which executes all the shutdown
hooks and releases the system resources. In the new code, however, the
resources are not released as the exception simply bubbles up to the main
method and thrown out, while the other threads are still running. So
the resources are not released and build.xml cannot continue processing
the HTML results with <junitreport ...>
The fix is simple, but the investigation was hard. Big thank you to Jacques.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1746737 13f79535-47bb-0310-9956-ffa450edef68
The patch that I provided only emptied the files when they should be
deleted. This removes them completely
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1746239 13f79535-47bb-0310-9956-ffa450edef68
This is another major refactoring exercise of the start
component in ofbiz. There are too many details in this commit
that may be summarized as follows:
- refactor Start.java to make it about 140 lines long and focused
only on high level function calls. Also minimized the number of
class fields (state) to only two (ServerState and Config)
- Separate the AdminPortThread into a new class with some refactoring
done to ensure interoperability with other classes
- delete CommonsDaemonStart as jsvc is not yet fully implemented in ofbiz
and it stands in the way of a proper refactoring of ofbiz
- refactor everything so that all dependencies between classes are explicit
through function calls (no hidden state or declarations)
- Move all the server processing logic (status, shutdown, start ...) to a
new class called StartupControlPanel. All methods in this class are static
with side effects only limited to the arguments passed to them.
- Add a few missing apache copyright headers
- Refactoring some method, variable and enum names to make them clearer
- Refactor Config.java and make it much cleaner and simpler by breaking
things down into private methods. Also changed the portoffset to be
a normal int instead of a boxed int (Integer)
- redefine most classes to be package protected (not public) to offer
more isolation from the rest of the framework
- redefine most classes to be final
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1745351 13f79535-47bb-0310-9956-ffa450edef68
as suggested by Jacques, the commons-cli screen is widened to be
80 characters instead of the default 74 characters. Also the test
command for java -jar ofbiz.jar is improved in terms of documentation
so that it does not break off into a newline which was confusing
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1744537 13f79535-47bb-0310-9956-ffa450edef68
Box the error messages with "=======" strings to highlight
the error the to user when calling java -jar ofbiz.jar --somewrongcommand
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1744525 13f79535-47bb-0310-9956-ffa450edef68