Unfortunately it seems I found a deadlock again, happened while testing content completion (repeatedly adding/removing a “.”, which is a completion trigger in this project) in Spoofax 0.6.0.2, Eclipse 3.6, JDK 1.6.0_23.

At some point it started to have short hangs (mixed 100%/no cpu usage) when modifying the text, and a little later Eclipse hung completely (no cpu usage) right after inserting/removing a character. Deadlock?

I got 5 stacktraces from jvisualvm; they are all nearly identical. I will attach one of them. I also got a heap dump if that is useful; I will keep it around for at least a few days.

I can reproduce partial hangs (10-30s) by quickly typing “.something” at a place where “.” triggers completion.

Submitted by Tobi Vollebregt on 21 January 2011 at 16:57
threaddump-1295622927749.tdump21 January 2011 at 16:58

On 22 January 2011 at 16:16 Lennart Kats commented:

This is a known issue :( Eclipse has content completion run in the UI thread which is just inherently unsafe for most non-UI operations. The “partial hangs” may be SWTSafeLock’s watch dog kicking in (which is really just a bad workaround).

Closed as a duplicate of Spoofax/252


On 22 January 2011 at 16:16 Lennart Kats closed this issue.

On 22 January 2011 at 16:16 Lennart Kats tagged duplicate

Log in to post comments