Segfault with 0.11.1 on julia 1.3

[benlorenz]

After switching from 0.11.0 to 0.11.1 julia 1.3 crashes when loading Polymake.jl, both in our CI and locally:

julia> versioninfo()
Julia Version 1.3.1
Commit 2d5741174c (2019-12-30 21:36 UTC)
Platform Info:
  OS: Linux (x86_64-pc-linux-gnu)
  CPU: AMD Ryzen 9 5900X 12-Core Processor
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-6.0.1 (ORCJIT, generic)
Environment:
  JULIA_DEPOT_PATH = /tmp/julia13dep

(julia13proj) pkg> st
    Status `/tmp/julia13proj/Project.toml`
  [1f15a43c] CxxWrap v0.11.1
  [d720cf60] Polymake v0.5.4

julia> using Polymake

Thread 1 "julia" received signal SIGABRT, Aborted.
__GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:49
49      ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.
(gdb) bt
#0  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:49
#1  0x00007ffff764a538 in __GI_abort () at abort.c:79
#2  0x00007ffff788c7ef in jl_intref (idx=<optimized out>, arr=<optimized out>) at /buildworker/worker/package_linux64/build/src/typemap.c:194
#3  mtcache_hash_lookup (offs=<optimized out>, tparam=<optimized out>, ty=<optimized out>, a=<optimized out>) at /buildworker/worker/package_linux64/build/src/typemap.c:255
#4  jl_typemap_intersection_visitor (map=<optimized out>, offs=offs@entry=0, closure=closure@entry=0x7fffcc4abe30) at /buildworker/worker/package_linux64/build/src/typemap.c:545
#5  0x00007ffff7884daa in jl_method_table_insert (mt=<optimized out>, method=method@entry=0x7fffc834ef80, simpletype=simpletype@entry=0x7fffeb4a1700 <jl_system_image_data+17472>)
    at /buildworker/worker/package_linux64/build/src/gf.c:1683
#6  0x00007ffff78b0319 in jl_insert_methods (list=0x7fffe6888e20, list=0x7fffe6888e20) at /buildworker/worker/package_linux64/build/src/dump.c:2225
#7  _jl_restore_incremental (f=f@entry=0x7fffcc4ac130, mod_array=mod_array@entry=0x7fffca812230) at /buildworker/worker/package_linux64/build/src/dump.c:3227
#8  0x00007ffff78b32c5 in jl_restore_incremental (fname=0x7fffe60d46a8 "/tmp/julia13dep/compiled/v1.3/Polymake/QizyK_tSM6m.ji", mod_array=0x7fffca812230)
    at /buildworker/worker/package_linux64/build/src/dump.c:3272
#9  0x00007fffea0a50d6 in japi1__include_from_serialized_3216 () at loading.jl:676
#10 0x00007fffea0a70c3 in japi1__require_search_from_serialized_2786 () at loading.jl:776
#11 0x00007fffea0b1782 in japi1__require_2174 () at loading.jl:1001
#12 0x00007fffea0b370d in japi1_require_2161 () at loading.jl:922
#13 0x00007fffea0f4c5c in japi1_require_4817 () at loading.jl:917
#14 0x00007ffff788779c in _jl_invoke (world=26767, mfunc=<optimized out>, nargs=2, args=0x7fffcc4ad6b8, F=0x7fffeeac1260 <jl_system_image_data+56770464>)
    at /buildworker/worker/package_linux64/build/src/gf.c:2135
#15 jl_apply_generic (F=0x7fffeeac1260 <jl_system_image_data+56770464>, args=args@entry=0x7fffcc4ad6b8, nargs=nargs@entry=2)
    at /buildworker/worker/package_linux64/build/src/gf.c:2305
#16 0x00007ffff78ba5ab in jl_apply (nargs=3, args=0x7fffcc4ad6b0) at /buildworker/worker/package_linux64/build/src/julia.h:1631
#17 call_require (var=0x7fffe5d956e8, mod=0x7fffeb4eac10 <jl_system_image_data+317776>) at /buildworker/worker/package_linux64/build/src/toplevel.c:399
#18 eval_import_path (where=where@entry=0x7fffeb4eac10 <jl_system_image_data+317776>, from=from@entry=0x0, name=name@entry=0x7fffcc4ad768, 
    keyword=keyword@entry=0x7ffff7a03a69 "using", args=<optimized out>, args=<optimized out>) at /buildworker/worker/package_linux64/build/src/toplevel.c:436
#19 0x00007ffff78bb9ea in jl_toplevel_eval_flex (m=m@entry=0x7fffeb4eac10 <jl_system_image_data+317776>, e=<optimized out>, fast=fast@entry=1, expanded=expanded@entry=0)
    at /buildworker/worker/package_linux64/build/src/toplevel.c:656
#20 0x00007ffff78bb77a in jl_toplevel_eval_flex (m=m@entry=0x7fffeb4eac10 <jl_system_image_data+317776>, e=e@entry=0x7fffe64102f0, fast=fast@entry=1, expanded=expanded@entry=0)
    at /buildworker/worker/package_linux64/build/src/toplevel.c:764
#21 0x00007ffff78bc82c in jl_toplevel_eval (m=m@entry=0x7fffeb4eac10 <jl_system_image_data+317776>, v=v@entry=0x7fffe64102f0)
    at /buildworker/worker/package_linux64/build/src/toplevel.c:823
#22 0x00007ffff78bc8b5 in jl_toplevel_eval_in (m=0x7fffeb4eac10 <jl_system_image_data+317776>, ex=0x7fffe64102f0) at /buildworker/worker/package_linux64/build/src/toplevel.c:843
#23 0x00007fffea0bcfd5 in japi1_eval_3749 () at boot.jl:330
#24 0x00007ffff788779c in _jl_invoke (world=26767, mfunc=<optimized out>, nargs=2, args=0x7fffcc4adab8, F=0x7fffed224310 <jl_system_image_data+30961744>)
    at /buildworker/worker/package_linux64/build/src/gf.c:2135
#25 jl_apply_generic (F=0x7fffed224310 <jl_system_image_data+30961744>, args=0x7fffcc4adab8, nargs=<optimized out>) at /buildworker/worker/package_linux64/build/src/gf.c:2305
#26 0x00007fffea2aa9ca in japi1_eval_user_input_11803 () at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.3/REPL/src/REPL.jl:86
#27 0x00007fffea2aacc5 in macro expansion () at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.3/REPL/src/REPL.jl:118
#28 japi1_#26_11801 () at task.jl:333
#29 0x00007ffff788779c in _jl_invoke (world=26767, mfunc=<optimized out>, nargs=0, args=0x7fffe8766eb0, F=0x7fffe879a9d0) at /buildworker/worker/package_linux64/build/src/gf.c:2135
#30 jl_apply_generic (F=0x7fffe879a9d0, args=args@entry=0x7fffe8766eb0, nargs=nargs@entry=0) at /buildworker/worker/package_linux64/build/src/gf.c:2305
#31 0x00007ffff78a2e79 in jl_apply (nargs=1, args=0x7fffe8766ea8) at /buildworker/worker/package_linux64/build/src/julia.h:1631
#32 start_task () at /buildworker/worker/package_linux64/build/src/task.c:659
#33 0x0000000000000000 in ?? ()

It works fine again when I switch back to 0.11.0.

I am having a hard time understanding that backtrace, bisecting points to a6be4ba. Any suggestions how to debug this further?

[benlorenz]

By chance we noticed that this also happens with 1.4.0 and 1.4.1, but not in 1.4.2. Looking at the very few changes between 1.4.1 and 1.4.2 I found this julia bug: JuliaLang/julia#29859 with a very similar backtrace that was fixed with JuliaLang/julia#35378 and backported for 1.4.2.

[barche]

OK, so anything in CxxWrap we can do about this, then? I don't see an obvious workaround.

[benlorenz]

I don't really know any details why this happens and was hoping for someone else to have an idea.
Somehow the cxxdereference change in a6be4ba triggers that julia bug and I wanted to dig in further to find the code from Polymake.jl where this happens but didn't have time for that so far.

[ufechner7]

Can this be closed now? Julia 1.3 is obsolete.