Aws::Xml::Parser::ParsingError when listing objects

[krystof-k]

Describe the bug

I get Aws::Xml::Parser::ParsingError error when listing objects using #list_objects_v2.

Regression Issue

• Select this option if this issue appears to be a regression.

Expected Behavior

It works.

Current Behavior

It throws the error:

response = s3_client.list_objects_v2(
  bucket:,
  prefix:,
  continuation_token:
)
/app/vendor/bundle/ruby/3.2.0/gems/aws-sdk-core-3.203.0/lib/aws-sdk-core/xml/parser/stack.rb:49:in `error': xmlParseCharRef: invalid xmlChar value 12 (Aws::Xml::Parser::ParsingError)

Reproduction Steps

Unable to figure out how to reproduce. It happens just for some prefixes, could be related to filenames.

Possible Solution

No response

Additional Information/Context

No response

Gem name ('aws-sdk', 'aws-sdk-resources' or service gems like 'aws-sdk-s3') and its version

aws-sdk-s3 1.160.0

Environment details (Version of Ruby, OS environment)

Ruby 3.2.5, Alpine 3.20

[alextwoods]

What XML gem are you using? (You can determine which xml "engine" is used at runtime with: Aws::Xml::Parser.engine, it should give something like Aws::Xml::Parser::OxEngine).

Any idea what prefixes might be causing an issue? Do you have any unusual object names in the bucket you're trying to list for? Does this happen everytime (accounting for paging) or are you able to successfully get all pages of the response sometimes?

[krystof-k]

Wow, that was fast :)

Aws::Xml::Parser.engine says Aws::Xml::Parser::NokogiriEngine.

It happens every time, but only for a certain page probably. Unfortunately I have no idea how to retrieve any further information, all I have is the continuation token.

And yes, it is possible there is some unusual character in the object name, it is happening for at least two files (unfortunately I don't know which ones).

[alextwoods]

I'm trying to reproduce this, but haven't been able to figure out what characters might be responsible yet. A few things that might help:

1. You can enable wire logs with the http_wire_trace option, eg: s3_client = Aws::S3::Client.new(http_wire_trace: true). That will log all of the raw xml response and should give us information about which keys are actually failing.
2. You can switch which xml engine you are using by installing another gem (eg include gem 'ox' in your bundlefile). (The order of preference they are selected by the SDK in is ox, oga, libxml, nokogiri and then rexml). Different xml parsers handle some characters differently.

[krystof-k]

Awesome, thanks. Now I can see that it is probably because of a corrupted filename. This would be the problematic object:

<Contents>
  <Key>
    room-tyj62m1bnw14/storage/files/2022/07/16/C1WECz5UQ8Xw82vzd7yZj8/\xEF\xBF\xBD&#x7;\xEF\xBF\xBDo&#xb;\xEF\xBF\xBDu\xEF\xBF\xBD]onf\xEF\xBF\xBDI\xDC\x99\xD6\xB3\xEF\xBF\xBD\xEF\xBF\xBD\xEF\xBF\xBDN.r\n\xEF\xBF\xBD</Key>
  <LastModified>2024-09-10T09:00:16.000Z</LastModified>
  <ETag>&quot;be2142edba71363ab371527b7865056c&quot;</ETag>
  <Size>86623</Size>
  <StorageClass>STANDARD</StorageClass>
</Contents>

However I can download the file (with the corrupted filename) from S3 console with no issue. So I'm not sure what is the right approach to this.

[alextwoods]

Can you try using a different XML gem? I believe there are differences in what the underlying xml parsers consider valid

[alextwoods]

May be related to #3081

[krystof-k]

Well, that will be a little bit complicated but I'll try.

[mullermp]

What specifically is complicated? You should just be able to add the 'ox' gem to your gemfile and the SDK will prefer it.

[krystof-k]

@mullermp Our production deployment pipeline :)

@alexwoods so I tried with Ox and it seems to work just fine, the response (object key) is the same.

[github-actions]

This issue is now closed. Comments on closed issues are hard for our team to see.
If you need more assistance, please open a new issue that references this one.

[krystof-k]

OK, so does this mean it is a "bug" in Nokogiri and the solution is to use another XML parser?

[mullermp]

I don't know if it's a bug or not but it's in a grey area for sure. Some parsers think it's invalid and others don't. I think nokogiri is just more strict on that specification. This isn't the first time we've seen this. S3 doesn't strictly follow xml rules either. In either case I would maybe consider adding validation to prevent file names having invalid characters.