[jira] Commented: (AXIS2-1121) binary and enumFacet are incompatible with unordered in ADBBeanTemplate
[
http://issues.apache.org/jira/browse/AXIS2-1121?page=comments#action_12433224 ]
Davanum Srinivas commented on AXIS2-1121:
-
Thanks Dims. I'm trying a build now. It takes a long time from here
due to all the maven library access delays (I already tried offline,
which failed hopefully only because a new library is requried).
The problems I've noted with the template remain. In the unordered case
the code inserts binary (or enumFacet) code into a looping structure
that looks for each property in any order. Each iteration of the loop
must check to see what property it is at and then load that value using
whatever method is correct. The code structure of the loop presumes
that each such property-loading code segment is an single if statement
that can be preceded by an else as there is no need to continue
checking properties in a single iteration once one is found. However,
the code generated for binary and enumFacet properties satisfies neither
of these criteria. This code does not check to see if it is at the
property it loads and it is not in the form of an if. This causes
invalid code to be generated. The binary and and enumFacet generated
code will only work in an ordered context.
With a binary MTOM property is it still possible to validate the
property by checking the name of the start element? If so, my fix
(which just backs out 434331 and 436663) will work. If not, then some
other test needs to be devised if binary attributes are to work in an
unordered context.
I'd be happy to fix if you could answer that question.
Thanks,
Chuck
binary and enumFacet are incompatible with unordered in ADBBeanTemplate
---
Key: AXIS2-1121
URL: http://issues.apache.org/jira/browse/AXIS2-1121
Project: Apache Axis 2.0 (Axis2)
Issue Type: Bug
Reporter: Davanum Srinivas
Hi Dims and Ajith,
I believe you guys added revisions 434331 and 436663 as part of a fix
for Axis2-1065. Unfortunately these changes break the build and the
template. E.g., the sforce enterprise wsdl generates bogus code and
kills a clean build.
Simply backing out the changes in these two revisions causes the build
to succeed and all tests to pass. However, I suspect this will regress
Axis2-1065 and the same type of problem exists for enumFacet.
The issue is this. For ordered property sets, the main property parsing
loop generates code like this (in pseudo-code notation):
if (elementName==name1) {
...
} else throw Unexpected subelement
if (elementName==name2) {
...
} else throw Unexpected subelement
...
While in the unordered case the same code looks like this:
while (!endElement) {
if (elementName==name1) {
...
}
else if (elementName==name2) {
...
}
...
}
The binary MTOM code that was added fills the role of one the embedded
if's but is not of the same form, not an if at all. This code works in
the first ordered case, but in the second unordered case it doesn't work
at all to have fixed code expecting to just read a binary value where
one of the if's is supposed to be. Also in the binary case in the
current template the else is still generated after the MTOM code that is
not an if, whence the syntax violation in the code generated for the
sforce wsdl, which evidently has a binary property in an unordered
property set.
I'm no expert on MTOM. Can't you still test that you've got the right
property (the binary property) by using the start-element name? For the
template to work in the unordered case, some test that you've got the
binary propery is essential. The existing template does not do this,
which can only work in the ordered case where you know you are at the
right position, and even in this case does not do the proper error
checking to verify this.
I would commit the attached template as axis2 passes all tests, but the
logs indicate you made the revisions that created this problem to fix
another problem and so I don't want to just back that out without
bringing this to your attention first.
Chuck
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
[jira] Commented: (AXIS2-1121) binary and enumFacet are incompatible with unordered in ADBBeanTemplate
[
http://issues.apache.org/jira/browse/AXIS2-1121?page=comments#action_12433257 ]
Chuck Williams commented on AXIS2-1121:
---
Dims,
Thanks for making this update.
I still have not managed to get a successful build (last attempt failed tests
in the security scenarios due to check-ins today), but have reviewed the
changes to the template.
I see that you explicitly did not check for being at the right property in the
unordered case (omitted the test of the start element name) and so presume that
for some reason this is not possible. This means the binary property code will
be evaluated on every iteration of the loop for any container with at least one
binary property. I don't see how this could work. In the sforce wsdl that
failed yesterday, it has an unordered type with both binary and non-binary
properties. The unordered property parsing loop can only work if it can detect
when it has reached the binary property. If the start element name cannot be
used for this, is there some other test that can?
I wrote the portions of the template that pertain to the unordered case and
understand it well, but I don't understand MTOM fully. I've read the spec and
don't see why testing the start element name cannot be done. If you could
explain how to test that the binary property has been reached I could fix the
the template.
binary and enumFacet are incompatible with unordered in ADBBeanTemplate
---
Key: AXIS2-1121
URL: http://issues.apache.org/jira/browse/AXIS2-1121
Project: Apache Axis 2.0 (Axis2)
Issue Type: Bug
Reporter: Davanum Srinivas
Hi Dims and Ajith,
I believe you guys added revisions 434331 and 436663 as part of a fix
for Axis2-1065. Unfortunately these changes break the build and the
template. E.g., the sforce enterprise wsdl generates bogus code and
kills a clean build.
Simply backing out the changes in these two revisions causes the build
to succeed and all tests to pass. However, I suspect this will regress
Axis2-1065 and the same type of problem exists for enumFacet.
The issue is this. For ordered property sets, the main property parsing
loop generates code like this (in pseudo-code notation):
if (elementName==name1) {
...
} else throw Unexpected subelement
if (elementName==name2) {
...
} else throw Unexpected subelement
...
While in the unordered case the same code looks like this:
while (!endElement) {
if (elementName==name1) {
...
}
else if (elementName==name2) {
...
}
...
}
The binary MTOM code that was added fills the role of one the embedded
if's but is not of the same form, not an if at all. This code works in
the first ordered case, but in the second unordered case it doesn't work
at all to have fixed code expecting to just read a binary value where
one of the if's is supposed to be. Also in the binary case in the
current template the else is still generated after the MTOM code that is
not an if, whence the syntax violation in the code generated for the
sforce wsdl, which evidently has a binary property in an unordered
property set.
I'm no expert on MTOM. Can't you still test that you've got the right
property (the binary property) by using the start-element name? For the
template to work in the unordered case, some test that you've got the
binary propery is essential. The existing template does not do this,
which can only work in the ordered case where you know you are at the
right position, and even in this case does not do the proper error
checking to verify this.
I would commit the attached template as axis2 passes all tests, but the
logs indicate you made the revisions that created this problem to fix
another problem and so I don't want to just back that out without
bringing this to your attention first.
Chuck
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
