3:14 p.m.
I didn't get any response in #wildfly, but I tested this and I found it to be true:
03:23:11 PM) mazz: I need a sanity check please. I'm trying to run unit tests on my
module subsystem (via my subclass of
org.jboss.as.subsystem.test.AbstractSubsystemBaseTest). My test DID work on WF8.2 but FAIL
when I move to WF9 CR2.
(03:23:11 PM) mazz: The error is :
(03:23:11 PM) mazz: failure description: "WFLYCTL0201: Unknown attribute
'org'"
(03:23:11 PM) mazz: in my custom subsystem, I have several attributes of the form
"org.abc.xyz" such as:
(03:23:11 PM) mazz: <element org.hawkular.package.name.foo="test1"
org.hawkular.another.package.blah-blah="test2" />
(03:23:44 PM) mazz: could it be possible that attribute names can no longer have
"." in their names?
(03:24:11 PM) mazz: I have no attributes called "org" but plenty that look like
"org.some.name.here"
(03:47:24 PM) mazz: uh-oh. Yeah, I thikn that's a problem
(03:47:48 PM) mazz: in WildFly 9 CR2, it is no longer possible to have a dot
(".") in attribute names.
(03:47:50 PM) mazz: This makes me sad
(03:48:06 PM) mazz: this was possible in WildFly 8.2.Final
(03:48:36 PM) mazz: I just changed all my "." to "-" in my custom
module subsystem extension (in the java code and the .xml) and now my tests pass
(03:50:06 PM) mazz: I don't know if that is also true for element names.
(03:50:25 PM) mazz: I have "org.abc.xyz" element names as well. but the error
message I got just mentioned attribute name
I suppose we could change these names so the "." is replaced with "-"
(they are also sysprop names that are used to replace ${key} tokens in bus config files,
so i used the standard "." notation for these names).
What should we do? Ask WF to fix this (it probably is a bug - I doubt they meant to make
that change) or just deal with it and workaround by changing our attribute names?
4:05 p.m.
Sounds like a pretty bad regression. They should likely fix it.
Although we may want also want to make a change to '-' instead of
waiting for a fix. Bad for sysprop names, though...
On 6/15/2015 4:14 PM, John Mazzitelli wrote:
I didn't get any response in #wildfly, but I tested this and I
found it to be true:
03:23:11 PM) mazz: I need a sanity check please. I'm trying to run unit tests on my
module subsystem (via my subclass of
org.jboss.as.subsystem.test.AbstractSubsystemBaseTest). My test DID work on WF8.2 but FAIL
when I move to WF9 CR2.
(03:23:11 PM) mazz: The error is :
(03:23:11 PM) mazz: failure description: "WFLYCTL0201: Unknown attribute
'org'"
(03:23:11 PM) mazz: in my custom subsystem, I have several attributes of the form
"org.abc.xyz" such as:
(03:23:11 PM) mazz: <element org.hawkular.package.name.foo="test1"
org.hawkular.another.package.blah-blah="test2" />
(03:23:44 PM) mazz: could it be possible that attribute names can no longer have
"." in their names?
(03:24:11 PM) mazz: I have no attributes called "org" but plenty that look like
"org.some.name.here"
(03:47:24 PM) mazz: uh-oh. Yeah, I thikn that's a problem
(03:47:48 PM) mazz: in WildFly 9 CR2, it is no longer possible to have a dot
(".") in attribute names.
(03:47:50 PM) mazz: This makes me sad
(03:48:06 PM) mazz: this was possible in WildFly 8.2.Final
(03:48:36 PM) mazz: I just changed all my "." to "-" in my custom
module subsystem extension (in the java code and the .xml) and now my tests pass
(03:50:06 PM) mazz: I don't know if that is also true for element names.
(03:50:25 PM) mazz: I have "org.abc.xyz" element names as well. but the error
message I got just mentioned attribute name
I suppose we could change these names so the "." is replaced with "-"
(they are also sysprop names that are used to replace ${key} tokens in bus config files,
so i used the standard "." notation for these names).
What should we do? Ask WF to fix this (it probably is a bug - I doubt they meant to make
that change) or just deal with it and workaround by changing our attribute names?
_______________________________________________
hawkular-dev mailing list
hawkular-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hawkular-dev
2:28 a.m.
On 15 Jun 2015, at 23:05, Jay Shaughnessy wrote:
Sounds like a pretty bad regression. They should likely fix it.
Although we may want also want to make a change to '-' instead of
waiting for a fix. Bad for sysprop names, though...
How exactly does that impact us?
Attribute names can't be FQDN, that I understand.
But why do we want them as FQDN in the first place?
I see one occurence in standalone.xml
<connector org.hawkular.bus.broker.connector.name="openwire"
org.hawkular.bus.broker.connector.protocol="tcp"
socket-binding="org.hawkular.bus.broker"/>
Can't that just be "protocol" ?
If WildFly forces a FQDN here for the protocol, then they need to
support it. Otherwise we can change our side.
Sysprops are in the form
<property name="bla" value="bla" />
and seem not affected
Heiko
8:32 a.m.
These aren't FQDN java names - these are sysprop names. They are like this because
these are actually used to "pass through" to the active mq broker configuration
file. I could name them something simpler, but the idea was you can definitely tell which
ones were sysprops to be passed through (and in the future we are going to need more of
these to further configure the bus broker).
This worked in 8.2. They didn't care what our names are of the attributes. Something
broke in 9.
----- Original Message -----
On 15 Jun 2015, at 23:05, Jay Shaughnessy wrote:
> Sounds like a pretty bad regression. They should likely fix it.
> Although we may want also want to make a change to '-' instead of
> waiting for a fix. Bad for sysprop names, though...
How exactly does that impact us?
Attribute names can't be FQDN, that I understand.
But why do we want them as FQDN in the first place?
I see one occurence in standalone.xml
<connector org.hawkular.bus.broker.connector.name="openwire"
org.hawkular.bus.broker.connector.protocol="tcp"
socket-binding="org.hawkular.bus.broker"/>
Can't that just be "protocol" ?
If WildFly forces a FQDN here for the protocol, then they need to
support it. Otherwise we can change our side.
Sysprops are in the form
<property name="bla" value="bla" />
and seem not affected
Heiko
_______________________________________________
hawkular-dev mailing list
hawkular-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hawkular-dev
9:07 a.m.
Tomas just answered on #wildfly:
<theute> BTW mazz found either an issue or a deliberate change in WF9. subsystem
attributes names used to accept things like "foo.bar" now it chokes on the
"."
<theute> ctomc, do you know who we can annoy to make sure that this is expected and
we need to change or if that's a bug and going to change back
<ctomc> theute i would be the one that you would need to annoy
<ctomc> but all i can tell you that this is done on purpose
<ctomc> "." is special charachers
<ctomc> that cannot be part of attribute name
<theute> :-/
<ctomc> as it is used for "enhaced syntax" resolving in cli/mgmt tools
<theute> and "-" is safe ?
<ctomc> yes
<ctomc> https://issues.jboss.org/browse/WFCORE-460
<jbossbot> jira [WFCORE-460] Extended sytnax for read/write-attribute [Resolved
(Done) Sub-task, Major, Tomaz Cerar] https://issues.jboss.org/browse/WFCORE-460
<ctomc> we added support for some neat stuff
<ctomc> https://gist.github.com/ctomc/91055a6f4e7502dcd130
<ctomc> on top of my head, forbiden chars are .[]() and few more
Thomas
On 06/16/2015 03:32 PM, John Mazzitelli wrote:
These aren't FQDN java names - these are sysprop names. They are
like this because these are actually used to "pass through" to the active mq
broker configuration file. I could name them something simpler, but the idea was you can
definitely tell which ones were sysprops to be passed through (and in the future we are
going to need more of these to further configure the bus broker).
This worked in 8.2. They didn't care what our names are of the attributes. Something
broke in 9.
----- Original Message -----
> On 15 Jun 2015, at 23:05, Jay Shaughnessy wrote:
>
>> Sounds like a pretty bad regression. They should likely fix it.
>> Although we may want also want to make a change to '-' instead of
>> waiting for a fix. Bad for sysprop names, though...
>
> How exactly does that impact us?
> Attribute names can't be FQDN, that I understand.
> But why do we want them as FQDN in the first place?
> I see one occurence in standalone.xml
>
> <connector org.hawkular.bus.broker.connector.name="openwire"
> org.hawkular.bus.broker.connector.protocol="tcp"
> socket-binding="org.hawkular.bus.broker"/>
>
> Can't that just be "protocol" ?
>
> If WildFly forces a FQDN here for the protocol, then they need to
> support it. Otherwise we can change our side.
>
> Sysprops are in the form
>
> <property name="bla" value="bla" />
>
> and seem not affected
>
> Heiko
> _______________________________________________
> hawkular-dev mailing list
> hawkular-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hawkular-dev
>
_______________________________________________
hawkular-dev mailing list
hawkular-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hawkular-dev
3481
days inactive
3482
days old
4 comments
4 participants
participants (4)
-
Heiko W.Rupp -
Jay Shaughnessy -
John Mazzitelli -
Thomas Heute