# StackOverflowError vs ulimit -s

**URL:** <https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395>\
**Category:** General Usage\
**Created:** [July 20, 2020, 6:53pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395 "2020-07-20T18:53:51Z")\
**Posts on this page:** 9\
**Page:** 1

<div class="post-metadata">

**Author:** ![recri](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/recri/32/5280_2.png) [@recri](https://discourse.julialang.org/u/recri)\
**Post date:** [July 20, 2020, 6:53pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/1 "2020-07-20T18:53:51Z")

</div>

I am computing a recursive function which overflows the stack for some argument values. Somewhere I heard that julia on linux uses the system stack size as set by `ulimit -s`. So why do these two computations, one with `ulimit -s` of 8192 and the other with `ulimit -s` 64 times larger, both fail with StackOverflowError and the exact same stack trace?

```julia
rec@t490:~$ ulimit -s
8192
rec@t490:~$ julia
               _depth
   _ _ _(_)_ | Documentation: https://docs.julialang.org
  (_) | (_) (_) |
   _ _ _| |_ __ _ | Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _ | |
  | | |_| | | | (_| | | Version 1.4.2 (2020-05-23)
 _/ |\ __'_|_|_|\__'_| | Official https://julialang.org/ release
|__/ |

julia> include("julia/fuse.jl")
computem (generic function with 1 method)

julia> m(big"2.96678",true)
ERROR: StackOverflowError:
Stacktrace:
 [1] BigFloat(; precision::Int64) at ./mpfr.jl:112
 [2] BigFloat at ./mpfr.jl:112 [inlined]
 [3] -(::BigFloat) at ./mpfr.jl:575
 [4] m(::BigFloat, ::Bool) at /home/rec/julia/fuse.jl:7
 [5] m(::BigFloat, ::Bool) at /home/rec/julia/fuse.jl:10 (repeats 9 times)
 [6] m(::BigFloat, ::Bool) at /home/rec/julia/fuse.jl:9
 [7] m(::BigFloat, ::Bool) at /home/rec/julia/fuse.jl:10 (repeats 10818 times)
 [8] top-level scope at REPL[2]:1

julia> 
rec@t490:~$ ulimit -s $(expr 256 \* 1024)
rec@t490:~$ ulimit -s
262144
rec@t490:~$ julia
               _
   _ _ _(_)_ | Documentation: https://docs.julialang.org
  (_) | (_) (_) |
   _ _ _| |_ __ _ | Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _ | |
  | | |_| | | | (_| | | Version 1.4.2 (2020-05-23)
 _/ |\ __'_|_|_|\__'_| | Official https://julialang.org/ release
|__/ |

julia> include("julia/fuse.jl")
computem (generic function with 1 method)

julia> m(big"2.96678",true)
ERROR: StackOverflowError:
Stacktrace:
 [1] BigFloat(; precision::Int64) at ./mpfr.jl:112
 [2] BigFloat at ./mpfr.jl:112 [inlined]
 [3] -(::BigFloat) at ./mpfr.jl:575
 [4] m(::BigFloat, ::Bool) at /home/rec/julia/fuse.jl:7
 [5] m(::BigFloat, ::Bool) at /home/rec/julia/fuse.jl:10 (repeats 9 times)
 [6] m(::BigFloat, ::Bool) at /home/rec/julia/fuse.jl:9
 [7] m(::BigFloat, ::Bool) at /home/rec/julia/fuse.jl:10 (repeats 10818 times)
 [8] top-level scope at REPL[2]:1

julia> 

```

---

<div class="post-metadata">

**Author:** ![Oscar\_Smith](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/oscar_smith/32/25343_2.png) [@Oscar\_Smith](https://discourse.julialang.org/u/Oscar_Smith)\
**Post date:** [July 20, 2020, 7:24pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/2 "2020-07-20T19:24:28Z")

</div>

Based on `(repeats 10818 times)` it looks like you probably have an infinite loop somewhere.

---

<div class="post-metadata">

**Author:** ![recri](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/recri/32/5280_2.png) [@recri](https://discourse.julialang.org/u/recri)\
**Post date:** [July 20, 2020, 9:04pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/3 "2020-07-20T21:04:58Z")

</div>

No loops, just recursion:

`m(x) = x < 0 ? -x : m(x-m(x-1))/2`

increasing the stack size with `ulimit -s` did not increase the depth of recursion reached when the stack overflowed.

---

<div class="post-metadata">

**Author:** ![recri](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/recri/32/5280_2.png) [@recri](https://discourse.julialang.org/u/recri)\
**Post date:** [July 25, 2020, 2:38am UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/4 "2020-07-25T02:38:59Z")

</div>

The solution is to use julia-1.5.0-rc1.

---

<div class="post-metadata">

**Author:** ![recri](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/recri/32/5280_2.png) [@recri](https://discourse.julialang.org/u/recri)\
**Post date:** [July 25, 2020, 2:58pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/5 "2020-07-25T14:58:16Z")

</div>

Oops, but only as a single threaded julia. If you start multiple threads they get 8Mbyte fixed stack segments. If you stay away from threads, then your julia will consume all the stack you can give it.

---

<div class="post-metadata">

**Author:** ![klaff](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/klaff/32/7637_2.png) [@klaff](https://discourse.julialang.org/u/klaff)\
**Post date:** [July 25, 2020, 3:27pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/6 "2020-07-25T15:27:19Z")

</div>

Do you know why this is the solution?

---

<div class="post-metadata">

**Author:** ![recri](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/recri/32/5280_2.png) [@recri](https://discourse.julialang.org/u/recri)\
**Post date:** [July 25, 2020, 6:23pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/7 "2020-07-25T18:23:53Z")

</div>

Nope, just stumbling around in the dark here. But I do have 16 copies of julia-1.5.0 computing values on an EC2 instance right now, the deepest completed evaluation in the group so far is 1066022.

---

<div class="post-metadata">

**Author:** ![klaff](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/klaff/32/7637_2.png) [@klaff](https://discourse.julialang.org/u/klaff)\
**Post date:** [July 25, 2020, 8:20pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/8 "2020-07-25T20:20:36Z")

</div>

Thank you, I was curious.

I’m no computer science expert, but I had reached the conclusion at some point (well before learning Julia) that recursion wasn’t a good idea if the number of recursions wasn’t bounded to be something reasonably small, and that if I felt I was headed in that direction, I should rewrite and make my own stacks as needed for the problem rather than relying on whatever happens under the covers of whatever language I was using.

However, I have not tried to learn about how Julia uses the stack.

---

<div class="post-metadata">

**Author:** ![recri](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/recri/32/5280_2.png) [@recri](https://discourse.julialang.org/u/recri)\
**Post date:** [July 26, 2020, 12:08pm UTC](https://discourse.julialang.org/t/stackoverflowerror-vs-ulimit-s/43395/9 "2020-07-26T12:08:23Z")

</div>

That’s generally good advice. In this case the variation in recursive depth of the function is itself interesting, and rewriting to replace the recursion with a different form of evaluation appears to be unnecessary.

At this point I’m merely curious to know whether the current behavior is intended and likely to persist.

And wondering how this core war is going to turn out, each julia process is working on progressively deeper recursions.

 ![Screenshot from 2020-07-26 07-59-00](https://global.discourse-cdn.com/julialang/original/3X/f/1/f1ecca1ff0750d7af3a93943968ee793f0d77adc.png)