Hi Taher,
Thanks for the fix, I have changed the log commit to refer to OFBIZ-9196
About your question below. Actually, the problem is not related with Buildbot particularly. It's obvious since you were able to reproduce locally
using only a trunk instance.
Now, since it was introduced between R15 and R16, the same problem exists in R16. But your fix can't be backported because things are handled
differently there.
I see 2 ways of fixing it
1) change the code in R16 in another way
2) (I guess) backport the changes which leaded to using the *.properties files
(like load-data.properties) and then backport your change
I'm inclined for the second
Jacques
Le 19/03/2017 à 14:15, Taher Alkhateeb a écrit :
Okay great!
Now the next question. Does buildbot fail on simultaneous commits to the
same branch or for different branches only? Meaning if we commit at the
same time to trunk it's okay but if we commit to trunk and 16.11 then it
fails?. I am asking because I assume only one build for one branch happens
at a time. Is this correct?
Cheers,
Taher Alkhateeb
On Mar 19, 2017 3:28 PM, "Jacques Le Roux" <jacques.le.r...@les7arts.com>
wrote:
Yes, that should do it
Jacques
Le 19/03/2017 à 13:01, Taher Alkhateeb a écrit :
Okay, so, to be absolutely exact, the repeat behavior is the following:
- checkout ofbiz r1744041
- ant clean-all load-demo
- ant start
- open a new terminal
- ant run-tests
- observe tests run successfully
And then
- checkout ofbiz r1744107
- ant clean-all load-demo
- ant start
- open a new terminal
- ant run-tests
- observe tests fail with the message Couldn't create server socket(/
127.0.0.1:10523) (Address already in use: JVM_Bind)Start.java
Is this correct?
On Sun, Mar 19, 2017 at 2:47 PM, Jacques Le Roux <
jacques.le.r...@les7arts.com> wrote:
yep, you got it. But you also need an OFBiz instance running (previously)
to reproduce, of course!
Jacques
Le 19/03/2017 à 12:37, Taher Alkhateeb a écrit :
Sorry, I'm still lost, trying to explain it again:
Before the commit: "ant clean-all load-demo run-tests" works
after the commit: "ant clean-all load-demo run-tests" does not work with
the socket error message?
Is this now what you mean?
On Sun, Mar 19, 2017 at 2:32 PM, Jacques Le Roux <
jacques.le.r...@les7arts.com> wrote:
It's quite simple, I repeat.
It's not specifically related with Buildbot despite the title. It was
1st
identified there.
If you run an OFBiz instance (I ran one at the "beforeGradle" tag, but
we
know any commit version fits since it's currently reproducible on
Buildbot)
Before your commit (r1744107 in disappeared trunk) below (ie using the
last previous commit, to be even more clear r1744041) "ant clean-all
load-demo run-tests" works
After (ie using r1744107) it fails with the error reported below. I
repeat
it here in case
------------------------------------------------------------
------------------------------------------------------------
------------------------------
run-tests:
[java] org.ofbiz.base.start.StartupException: Couldn't create
server
socket(/127.0.0.1:10523) (Address already in use: JVM_Bind)Start.java
using configuration file org/ofbiz/base/start/test.properties
------------------------------------------------------------
------------------------------------------------------------
------------------------------
I hope it's now clear. And yes it's in the commit and we need to find
where!
You asked me to answer on HipChat, sorry I'll not, I must move now...
HTH
Jacques
Le 19/03/2017 à 12:00, Taher Alkhateeb a écrit :
The commit in question is not important because there are too many
changes.
What we need is a "repeat" behavior. So my questions from this thread
remain to be answered before we can investigate.
On Sun, Mar 19, 2017 at 1:53 PM, Jacques Le Roux <
jacques.le.r...@les7arts.com> wrote:
Here it is
------------------------------------------------------------
------------------------------------------------------------
------------------------------
git.exe bisect good
298c831a3aa1cea45b4aab66acebd9617f6615e0 is the first bad commit
commit 298c831a3aa1cea45b4aab66acebd9617f6615e0
Author: Taher A. Alkhateeb <ta...@apache.org>
Date: Mon May 16 18:40:49 2016 +0000
major change in the start component related to OFBIZ-6783
this is a big commit that achieves the following new features in
ofbiz:
- add the commons-cli library
- fix build.xml in start menu to include the commons-cli. It is done
in a way to ensure that the classpath continues to work when
ofbiz.jar
is copied to the root folder
- set a default portoffset value of 0 when not selected in build.xml
- fully redefine the commands passed in java -jar ofbiz.jar using the
commons-cli
api. The commands are now much more consistent and clean
- remove ofbiz dependency on String[] args across the different
components
and isolate
them in a new entity class called StartupCommand. This decouples
ofbiz
from both
String[] args and commons-cli at the same time
- fix all the calls in the master build.xml to comply with the new
commands
- fix the Config.java to remove dependecies on the args array
- create a utility class StartupCommandUtil that takes care of all
commons-cli
operations and abstracts away the implementation in private methods
- substantially reduce the size of main and init in Start.java by
refactoring
them in different places
- create an intermediate function called populateLoaderArgs. This is
the
first
step in removing the dependecies on args by "adapting" them using
this
method
- unify the exception model everywhere to StartupException. This
makes
exception
propagation cleaner and easier
- lots of cleanup of the code related in all these areas
Big thanks to Jacques for the substantial help in testing.
git-svn-id: https://svn.apache.org/repos/asf/ofbiz/trunk@1744107
13f79535-47bb-0310-9956-ffa450edef68
:100644 100644 e7ffd6978ec0b072e73fa128ec9e2e83ba81524a
74787e5942a4d0bae826bb06c15a19e24066fc30 M .classpath
:100644 100644 042fda025e68f8f86c359abe70b87b7138c31133
5053165b4a75e71d9f63768a84ff5351afe783a5 M build.xml
:040000 040000 31fa9e03d92cc7629cba86898c34da77a7a88ace
bb1541c40aa7f92ef5b4f125dc415eb394e1f93f M framework
Success (422 ms @ 19/03/2017 11:51:29)
------------------------------------------------------------
------------------------------------------------------------
------------------------------
Now we need to find where it is exactly...
Jacques
Le 19/03/2017 à 10:53, Taher Alkhateeb a écrit :
Hi Jacques,
Sorry I still don't understand, let me try to capture this.
Scenario 1 - Using ant
- Start OFBiz using: ant start
- Run tests in another OFBiz using: ant run-tests
- OFBiz fails with an error message
Scenario 2 - Using gradle
- Start OFBiz using: ./gradlew ofbiz
- Run tests in another OFBiz using: ./gradlew "ofbiz --test"
- OFBiz freezes?
Is this correct? Is this what you are witnessing?
On Sun, Mar 19, 2017 at 12:06 PM, Jacques Le Roux <
jacques.le.r...@les7arts.com> wrote:
Hi Taher,
Le 19/03/2017 à 04:18, Taher Alkhateeb a écrit :
Hi Jacques,
Okay just to try and understand your point, what are the
circumstances
in
which you are witnessing failure in post gradle but not
pre-gradle?
Fortunately it's now simple.
You run an OFBiz instance with "ant start"
You run tests with the same (or another) OFBiz instance with "ant
run-tests", bingo the port 10523 is already used tests stop with
the
error
below
In other words what is the repeat process, the expected outcome and
actual
outcome. I ask because I'm a bit confused not sure if you're
referring
to
web ports or admin ports pr something else?
Any port used by OFBiz I guess. But the tests fails on the the 1st
called
port: 10523 by default.
We got issues with the port 8080 on Buildbot so it was confusing.
Those
were due to Buildbot global config (not specific to OFBiz Buildbot
config)
ans has been fixed by Infra.
Jacques
Cheers,
Taher Alkhateeb
On Mar 18, 2017 10:24 PM, "Jacques Le Roux" <
jacques.le.r...@les7arts.com
wrote:
OK, I locally kept a version of the trunk just prior the Gradle
switch.
When running an OFBiz instance (any one fits, I used the
ofbiz-framework+plugin HEAD) you can run tests with R15, but not
with
this
preGradle version. You get
run-tests:
[java] org.ofbiz.base.start.StartupException: Couldn't
create
server
socket(/127.0.0.1:10523) (Address already in use:
JVM_Bind)Start.java
using configuration file org/ofbiz/base/start/test.properties
So the problem is anterior the Gradle switch and after the R15
freezing.
But unfortunately we no longer have a trunk to get back into
commits.
Fortunately we have https://github.com/apache/ofbiz
I'll try that :)
Jacques
Le 18/03/2017 à 17:43, Jacques Le Roux a écrit :
Le 18/03/2017 à 17:07, Taher Alkhateeb a écrit :
I think we discussed this in the past, but I will ask again for
clarity
...
what does gradle have to do with port blocking?
For now I have no clear evidences. Only that before the Gradle
move
we
had not the ports issues (locally and on Buildbot) I describe
in
this
Jira.
We need to dig deeper in this and localise the commit which
introduced
this
error.
In the meantime better not committing in parallel in the trunk
and
R16
if
we want to avoid the ports conflicts on Buildbot
How can gradle or any build tool be responsible for blocking
ports
that
are
not related to the build system but rather used from OFBiz
directly?
It could be code changes while refactoring, still a
supposition.
You
spoke about Config.java, at 1st glance I saw nothing there.
Jacques
On Mar 18, 2017 7:00 PM, "Jacques Le Roux" <
jacques.le.r...@les7arts.com
wrote:
It's clearly related with OFBIZ-9196 "Regression: the
testIntegration
Gradle taks shoud not use/block the ports"
The build in trunk and R16 overlap in time:
https://ci.apache.org/builders/ofbiz-trunk-framework/builds/
44/
Start Sat Mar 18 09:25:12 2017
End Sat Mar 18 09:26:56 2017
https://ci.apache.org/builders/ofbiz-branch16/builds/17
Start Sat Mar 18 09:25:30 2017
End Sat Mar 18 09:30:44 2017
Jacques
Le 18/03/2017 à 15:01, Taher Alkhateeb a écrit :
The problem after a second review of the logs is pretty clear.
You
have
two
instances of OFBiz running and conflicting on ports. Either
the
first
build
did not terminate while the second starts or shutdown is not
occurring
cleanly. In other words, if you start ./gradlew "ofbiz
load-data"
&&
./gradlew "ofbiz load-data" then it would work correctly
because
the
first
scripts terminates with exit code 0 before the second one
begins.
Most likely a buildbot script issue.
On Sat, Mar 18, 2017 at 4:24 PM, Taher Alkhateeb <
slidingfilame...@gmail.com
wrote:
Ctrl-C does not kill gradle and gradle does not need time to
release
resources
On Mar 18, 2017 4:21 PM, "Jacques Le Roux" <
jacques.le.r...@les7arts.com
wrote:
Thanks James,
Yes indeed, it takes a bit for Gradle, and especially its
wrapper,
to
disconnect resources. I noticed that too on Windows, thought
it
was
specific to Windows .
This could be the reason, though we don't use the wrapper
in
Buildbot.
We
could try to add a small pause (say 30 seconds) before
launching
a
build.
Because we (especially me) sometimes launch Buildbot builds
in
burst
when
backporting.
But it also happens for isolated build so I fear it's not
enough
Jacques
Le 18/03/2017 à 13:41, james yong a écrit :
I got this issue too when I pressed CTL-C to exit OFBiz and
then
ran
"./gradew ofbiz" immediately.
Jacques Le Roux wrote
Mmm, this is pretty bad
https://ci.apache.org/builders
/ofbiz-trunk-framework/builds/
44/steps/shell_1/logs/stdio
Building 84% > :ofbiz --load-dataSet OFBIZ_HOME to -
/home/buildslave32/slave32/ofbiz-trunk-framework/build
org.apache.ofbiz.base.start.StartupException: Couldn't
create
server
socket(/127.0.0.1:10523) (Address already in use (Bind
failed))
at
org.apache.ofbiz.base.start.AdminServer.
<init>
(AdminServer.java:56) Something is
wrong with Buildbot. I already asked infra about several
instances
running
at the same time, I don't know much yet... Jacques
Le 18/03/2017 à 10:26,
buildbot@
a écrit :
The Buildbot has detected a build exception on builder
ofbiz-trunk-framework while building . Full details are
available
at:
https://ci.apache.org/builders/ofbiz-trunk-framework/
builds/44
Buildbot URL: https://ci.apache.org/
Buildslave for this Build: silvanus_ubuntu
Build Reason: The AnyBranchScheduler scheduler named
'on-ofbiz-framework-commit' triggered this build
Build Source Stamp: [branch ofbiz/ofbiz-framework/trunk]
1787535
Blamelist: jleroux
BUILD FAILED: exception shell_1 upload_1
Sincerely,
-The Buildbot
--
View this message in context:
http://ofbiz.135035.n4.nabble
.
com/buildbot-exception-in-on-o
fbiz-trunk-framework-tp4703602
p4703620.html
Sent from the OFBiz - Dev mailing list archive at
Nabble.com.
