EC2 DescribeInstances API doesn't work (and likely many others)

[steveniemitz]

I was attempting to use EC2Client::DescribeInstances, but never got any reservations back.

Digging into the code, it looks like the generated parsers are incorrect. DescribeInstancesResult looks for a "Reservations" XML node, however the actual XML response is "reservationsSet". It looks like the generated code isn't looking for the locationName as specified in the API JSON descriptors, and instead only using the member key.

[steveniemitz]

It looks like this patch to the code generator template was enough to fix most places:

diff --git a/code-generation/generator/src/main/resources/com/amazonaws/util/awsclientgenerator/velocity/cpp/xml/ModelClassMembersDeserializeXml.vm b/code-generation/generator/src/main/resources/com/amazonaws/util/awsclientgenerator/velocity/cpp/xml/ModelClassMembersDeserializeXml.vm
index 6208d31..b01a4c7 100644
--- a/code-generation/generator/src/main/resources/com/amazonaws/util/awsclientgenerator/velocity/cpp/xml/ModelClassMembersDeserializeXml.vm
+++ b/code-generation/generator/src/main/resources/com/amazonaws/util/awsclientgenerator/velocity/cpp/xml/ModelClassMembersDeserializeXml.vm
@@ -29,7 +29,11 @@
 #set($listVarName = $CppViewHelper.computeVariableName($memberName) + "Member")
 #set($listMemberName = "member")
 #end##location specified in model
+#if($member.locationName)##location specified
+    XmlNode ${lowerCaseVarName}Node = resultNode.FirstChild("${member.locationName}");
+#else##no location specified
     XmlNode ${lowerCaseVarName}Node = resultNode.FirstChild("${memberName}");
+#end##

I'll submit a PR soon. Based on the diff, it looks like this mostly impacted EC2 APIs, but there were some changes in S3 as well.

[steveniemitz]

Looks like there's a similar problem on the request side as well. For instance the InstanceIds filter is serialized incorrectly.

[JonathanHenson]

thanks for your work on this, after you submit the pull request, I'll rerun the integration tests and make sure we didn't break anything. Ec2 is a special snowflake when it comes to protocol and is unfortunately an edge case we haven't tested too much.

[steveniemitz]

Cool, will do. How do you prefer the commits for this? Should I do one for the code-gen change and one with the new cpp files, or just all at once?

[JonathanHenson]

just the code-gen change. I'll regenerate clients on my end. Once we've verified the fix, we'll regenerate all clients and push them out.

[steveniemitz]

Any luck on this / the pull request?

[JonathanHenson]

Im testing it now. Should finish today some time.
On Jan 13, 2016 7:03 AM, "Steven Niemitz" notifications@github.com wrote:

Any luck on this / the pull request?

—
Reply to this email directly or view it on GitHub
#70 (comment).

[steveniemitz]

Great! I'll probably have another PR coming up too to fix the filter serialization.

[steveniemitz]

I submitted a preliminary PR for fixing filters, but I'm pretty sure its not 100% correct. I'd appreciate if you could check it out at some point and comment on it? Thanks!

#80

[steveniemitz]

ping?

[JonathanHenson]

so I merged in this pull request, but your other one is almost correct (though breaks SNS). I'm going to regenerate everything and push the new clients up. If you can look at my comments on the other PR, I can get that merged in as soon as it is ready.

[JonathanHenson]

just pushed the updates along with fixes for SNS