Instrument Error, Lock Critical Section Overlapping.

[lmnsyunhao]

Hello, I found some problem about wiretap. Here is a trace that instrumented by latest version of master. Line 386-387 seems two thread hold a some lock at the same time.

   375  t1.0178 read r!a.buffer r!1
   376  t1.0179 read r!a.name r!b
   377  t1.0180 enter r!1 boundedbuffer/Buffer.enq:(Ljava/lang/Object;)V
   378  t1.0181 request r!1
   379  t1.0182 acquire r!1
   380  t1.0183 read r!1._last i32!0
   381  t1.0184 read r!1._BUFSIZE i32!4
   382  t1.0185 read r!1._first i32!1
   383  t1.0186 branch
   384  t1.0187 release r!1
   385  t1.0188 request r!1
   386  t1.0189 acquire r!1
   387  t3.0008 acquire r!1
   388  t3.0009 read r!1._last i32!0
   389  t3.0010 read r!1._BUFSIZE i32!4
   390  t3.0011 read r!1._first i32!1
   391  t3.0012 branch
   392  t3.0013 release r!1
   393  t3.0014 request r!1
   394  t3.0015 acquire r!1
   395  t2.0058 acquire r!1
   396  t2.0059 read r!1._first i32!1

Thanks for solving it!

[kalhauge]

Hi! Great of you to reach out.

Can you send me the full trace, the program you ran it on and the commands you ran, that way I can try to reproduce the bug.

I’m a little buzzy this weekend, so I might first get to it next week :)

[lmnsyunhao]

Sorry, When I trying to reproduce this bug 10 times. But it's not happened again. I'm also confused. Maybe, it's not a bug? Sorry for bothering.

[kalhauge]

No problem, it's not a bother. Parallel bugs are inherently hard to reproduce, so I'm going to leave the issue open for now.

It it happens again remember to store the files and the commands used, then I can try to debug the program. If you set the
-Dwiretap.classfilesfolder=_wiretap/classes it stores the instrumented classes. This can be useful for debugging.