[ https://issues.apache.org/jira/browse/TIKA-2785?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16693319#comment-16693319 ]
Tim Allison commented on TIKA-2785: ----------------------------------- Sorry, you're right, [~lfcnassif]. Thank you for raising this. I just tested the ForkParser with a log file configured to write to stdout, and I tested it with the updated MockParser that grabs stdout statically, and the ForkParser worked well with both. > Switch parent/child IPC to mmap file from stdout/stderr in tika-server > ---------------------------------------------------------------------- > > Key: TIKA-2785 > URL: https://issues.apache.org/jira/browse/TIKA-2785 > Project: Tika > Issue Type: Task > Reporter: Tim Allison > Assignee: Tim Allison > Priority: Major > Fix For: 2.0.0, 1.20 > > > When I added {{-spawnChild}} to tika-server, I followed the pattern in > ForkParser of redirecting stdout/stderr for the parent to listen to output > from the child process. This relies on the parsers not grabbing a pointer to > stdout/stderr _before_ we redirect it. However, it is possible for parsers > to grab stdout/stderr during static initialization, as happens with log4j > configured to write to stdout/stderr. > This means that the communication channel from child to parent is unreliable. > Let's switch child to parent communication in {{-spawnChild}} to a shared > mmap'd file. -- This message was sent by Atlassian JIRA (v7.6.3#76005)