Fixed in SVN. Fix will be available in next Sesame release.

Turns out the code from QueryParserUtil.removeSPARQLQueryProlog is duplicated in the HTTPRepositoryConnection class: it has its own private method that performs the same task (and thus has the same regex bug).

Turns out the fix was only half the problem. My bad for not testing the actual workbench.

root cause

org.openrdf.repository.RepositoryException: Server responded with an unsupported file format: application/sparql-results+xml
org.openrdf.http.client.HTTPClient.getRDF(HTTPClient.java:1166)
org.openrdf.http.client.HTTPClient.sendGraphQuery(HTTPClient.java:517)
org.openrdf.http.client.HTTPClient.sendGraphQuery(HTTPClient.java:492)
org.openrdf.repository.http.HTTPGraphQuery.evaluate(HTTPGraphQuery.java:41)
org.openrdf.workbench.commands.QueryServlet.evaluateGraphQuery(QueryServlet.java:231)
org.openrdf.workbench.commands.QueryServlet.service(QueryServlet.java:152)

Hi Jeen,
the problem seems to not be solved. I've just tested the versione 2.6.3 and it continues throwing a "Server responded with an unsupported file format: application/sparql-results+xml"

Issue is caused by a bug in the regular expression that strips off the base declaration, in QueryParserUtil.removeSPARQLQueryProlog. Regex is too greedy, matching with everything in the query ending with "base>". This causes the query parser to throw an error.

Confirmed, and is not specific to JBoss, also happens in Tomcat 6.0.