From Bugzilla Helper: User-Agent: Mozilla/5.0 Galeon/1.2.7 (X11; Linux i686; U;) Gecko/20030131 Description of problem: when attempting to connect to a cvs archive in the cvs perspective, connecting to a pserver enabled cvs server stalls, and fails to connect Version-Release number of selected component (if applicable): beehive-compiled; maybe others How reproducible: Always Steps to Reproduce: 1. Attempt connect via CVS perpsective to pserver cvs server 2. 3. Actual Results: The connection will stall or fail, and sometimes eclipse will crash. ext servers work fine Expected Results: The naviagtor view should have been populated with a list of projects Additional info: First reported by ahussain, then by agreen. Entered by pmuldoon
I tried this today. I used this repository: :pserver:anoncvs.org:/cvsroot/monotone I checked out monotone just fine. It is a smallish repository. Can you tell me how you got this to fail?
It failed off 2.0.1 system - so it might have been fixed in the 2.1 version. According to ahussain and agreen, the reporters, any pserver respository would just fail to connect after an extended connection attempt. If it works in the gcj version of 2.1 it may have been fixed through upgrading our sources.
Ok, I'm closing this as fixed.
Created attachment 93091 [details] .log trace I'm still seeing a problem with eclipse 2.1, gcj-compiled. Try the following repository :pserver:anonymous.org:/home/eclipse Behavior is the same as Phil described in the initial report.
I've tried with more repositories (including both eclipse and one on sourceforge) and still can't reproduce. The stack trace you attached is suggestive. This appears to be a "shouldn't happen" situation. Unless the stack trace is incorrect (possible), the only possible NullPointerException in PlainSocketImpl$SocketInputStream.read would occur if this$0 was null. But in theory that cannot be the case. It might be informative to turn on debugging. I think you can do this by making a file called ".options" (anyplace is fine -- just run eclipse from that directory). In .options, put: org.eclipse.team.core/streams = true Then run eclipse with the "-debug" option. If this works, eclipse will print debug info when it is connecting to the cvs server. If you could attach this log from a failing run, that may give me the info I need. If that fails, the next step would be for me to try to run it from Toronto and debug a failure live. I really hope to avoid that...
Sigh. Somehow my libgcj-ssa RPM was left out of my gcc-ssa update a while back, and consequently, I've been building and testing with libgcj-ssa-3.5ssa-0.20030428.17. I updated to libgcj-ssa-3.5ssa-0.20030617.24, and this bug is gone. Closing.