Hi, I agree that an additional safeguard for what is deemed "empty" could be added here. If you like to attempt a Pull Request, feel free. Or if not, perhaps create a JIRA issue.
Jan > 9. nov. 2024 kl. 20:45 skrev Dmitri Maziuk <dmitri.maz...@gmail.com>: > > On 11/7/24 04:53, David Eric Pugh wrote: >> Would you mind testing this in either 9.7 or main? Make sure that we don't >> have a bug that needs fixing? > > @David, > > In case you haven't followed the thread: for some reason there is the 0x00 > byte where Solr expects "no data or a JSON object" (quoting Jan); it's only > in some configsets and we have no idea how it got there. > > Whether it's a bug or not, and whose, it may be worth adding a check before > passing the value to JSON parser: the "\0" is "no data"/empty string, at > least in traditional C definition. The trace shows it trying to parse the > string "\0" (both dots below are 00 in hex): > > org.noggit.JSONParser$ParseException: JSON Parse Error: char=.,position=0 > AFTER='.' BEFORE='' => org.noggit.JSONParser > $ParseException: JSON Parse Error: char=.,position=0 AFTER='.' BEFORE='' > > That is passed from > org.apache.solr.handler.admin.ConfigSetsHandler.handleConfigUploadRequest(ConfigSetsHandler.java:205) > to > org.apache.solr.common.util.Utils.fromJSON(Utils.java:305) > > It's probably a good idea to check if the string is zero-length somewhere > between those two. > > Dima >
