missing backtrace after task switch

[damiendr]

Normal backtrace:

# file test.jl
try
    error("test")
catch e
    rethrow(e)
end

$ julia test.jl
ERROR: LoadError: test
 in test at test.jl:4
 in include at ./boot.jl:261
 in include_from_node1 at ./loading.jl:392
 in process_options at ./client.jl:277
 in _start at ./client.jl:377
while loading test.jl, in expression starting on line 10

Missing backtrace when using print in the catch block:

# file testprint.jl
try
    error("test")
catch e
    println("error")
    rethrow(e)
end

$ julia testprint.jl
error
ERROR: LoadError: test
while loading testprint.jl, in expression starting on line 5

Works fine when run from the REPL:

julia> include("testprint.jl")
error
ERROR: LoadError: test
 in error at ./error.jl:21
 in include at ./boot.jl:261
 in include_from_node1 at ./loading.jl:392
 in eval at ./boot.jl:264
while loading testprint.jl, in expression starting on line 5

Happens both in Julia 0.4 and the latest 0.5-dev.

[JobJob]

What OS are you using? I couldn't replicate on OSX

[damiendr]

I get this behaviour on the following platforms (Note: the linux machine was accessed from the mac via ssh):

OSX 10.9.5
Julia Version 0.5.0-dev+1231
Commit f174d12* (2015-11-10 19:27 UTC)
Platform Info:
  System: Darwin (x86_64-apple-darwin13.4.0)
  CPU: Intel(R) Core(TM) i7-2635QM CPU @ 2.00GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Sandybridge)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.3

Linux Mint (3.13.0)
Julia Version 0.4.0
Commit 0ff703b* (2015-10-08 06:20 UTC)
Platform Info:
  System: Linux (x86_64-linux-gnu)
  CPU: Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz
  WORD_SIZE: 64
  BLAS: libopenblas (NO_LAPACK NO_LAPACKE DYNAMIC_ARCH NO_AFFINITY Haswell)
  LAPACK: liblapack.so.3
  LIBM: libopenlibm
  LLVM: libLLVM-3.3

It seems to be dependent on previous writes: adding println("hello") before the try...catch block in my example makes the backtrace appear. Conversely I've had backtraces disappear in situations where there were things printed before. Maybe a missing flush() or something?

[damiendr]

In fact even adding println("") at the start makes the backtrace appear.

[JobJob]

Ok so I had some printing happening in my .juliarc.jl, when I disabled that I'm able to reproduce (on OSX) using this method above:

# file testprint.jl
try
    error("test")
catch e
    println("error")
    rethrow(e)
end

and similarly any printing before the try brings the backtrace back.

[damiendr]

Also happens in cells of the IJulia notebook, one cell counting as one file of the above examples. Which is how I came across this bug, since cells are more likely than whole files to meet the conditions that trigger the bug.

[KristofferC]

Closing in favour of #19979.