I just committed a minor match suggested by Jim Murphy in SOLR-42 to slightly lower the safe read ahead limit to avoid reading beyond a a mark. Could you try out trunk (or wait until the next nightly build?)
-Yonik http://www.lucidimagination.com On Tue, May 12, 2009 at 10:57 AM, Nikolai Derzhak <niko...@zapatec.net> wrote: > OK. I've applied dirty hack as temporary solution: > > in src/java/org/apache/solr/analysis/HTMLStripReader.java of 1.4-dev - > enclosed io.reset in try structure. > > ( * @version $Id: HTMLStripReader.java 646799 2008-04-10 13:36:23Z yonik $) > " > private void restoreState() throws IOException { > try { > in.reset(); > } catch (Exception e) { > } > pushed.setLength(0); > } > > " > > But how to resolve this problem more civilized ? > > On Tue, May 12, 2009 at 12:20 PM, Nikolai Derzhak <niko...@zapatec.net>wrote: > >> Good day, people. >> >> We use solr to search in mailboxes (dovecot). >> But with some "bad" messages solr 1.4-dev generate error: >> " >> SEVERE: java.io.IOException: Mark invalid >> at java.io.BufferedReader.reset(BufferedReader.java:485) >> at >> org.apache.solr.analysis.HTMLStripReader.restoreState(HTMLStripReader.java:171 >> >> ..... >> " >> It's issue known as SOLR-42. >> >> How i can log field stored in index (i need message uid) ? >> >> How to ignore such error and/or message ? >> >> Thanks >
