[
https://issues.apache.org/jira/browse/SOLR-1091?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12748212#action_12748212
]
Yonik Seeley commented on SOLR-1091:
------------------------------------
Hmmm, I thought this might be a decoding error. Instead it perhaps looks like
the JVM not handling UTF8 encoding the same as Jetty.
As one can see by this:
$ curl --trace -
'http://localhost:8983/solr/select?q=abcdef&a=%ED%AF%80%ED%B1%B8&wt=phps'
The ascii portion of the dump looks like:
s:4:"......";
So the JVM is telling us that the string will serialize to 4 bytes, and when
Jetty actually does the serialization, it comes out to 6.
Java's String.getBytes() does have the following warning:
* <p> The behavior of this method when this string cannot be encoded in
* the given charset is unspecified.
> "phps" (serialized PHP) writer produces invalid output
> ------------------------------------------------------
>
> Key: SOLR-1091
> URL: https://issues.apache.org/jira/browse/SOLR-1091
> Project: Solr
> Issue Type: Bug
> Components: search
> Affects Versions: 1.3
> Environment: Sun JRE 1.6.0 on Centos 5
> Reporter: frank farmer
> Priority: Minor
> Fix For: 1.4
>
>
> The serialized PHP output writer can outputs invalid string lengths for
> certain (unusual) input values. Specifically, I had a document containing
> the following 6 byte character sequence: \xED\xAF\x80\xED\xB1\xB8
> I was able to create a document in the index containing this value without
> issue; however, when fetching the document back out using the serialized PHP
> writer, it returns a string like the following:
> s:4:"􀁸";
> Note that the string length specified is 4, while the string is actually 6
> bytes long.
> When using PHP's native serialize() function, it correctly sets the length to
> 6:
> # php -r 'var_dump(serialize("\xED\xAF\x80\xED\xB1\xB8"));'
> string(13) "s:6:"􀁸";"
> The "wt=php" writer, which produces output to be parsed with eval(), doesn't
> have any trouble with this string.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
