Compiler segfault in alpine but not centos8

[will]

On centos8, everything works alright:

[centos@ip-172-28-240-100 cb]$ uname -a
Linux ip-172-28-240-100.us-west-2.compute.internal 4.18.0-305.3.1.el8.x86_64 #1 SMP Tue Jun 1 16:14:33 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux
[centos@ip-172-28-240-100 cb]$ crystal --version
Crystal 1.1.1 [6d9a1d583] (2021-07-26)

LLVM: 10.0.1
Default target: x86_64-unknown-linux-gnu
[centos@ip-172-28-240-100 cb]$ shards build -s
Dependencies are satisfied
Building: cb
Parse:                             00:00:00.002276749 (   1.00MB)
Semantic (top level):              00:00:00.573018851 (  74.32MB)
Semantic (new):                    00:00:00.002739744 (  74.32MB)
Semantic (type declarations):      00:00:00.100629685 (  74.32MB)
Semantic (abstract def check):     00:00:00.009171972 (  74.32MB)
Semantic (ivars initializers):     00:00:00.021925261 (  90.32MB)
Semantic (cvars initializers):     00:00:00.068884934 ( 106.32MB)
Semantic (main):                   00:00:01.934609761 ( 250.75MB)
Semantic (cleanup):                00:00:00.006690541 ( 250.75MB)
Semantic (recursive struct check): 00:00:00.001527295 ( 250.75MB)
Codegen (crystal):                 00:00:02.820418446 ( 299.25MB)
Codegen (bc+obj):                  00:00:01.468691736 ( 299.25MB)
Codegen (linking):                 00:00:00.642008240 ( 299.25MB)

Codegen (bc+obj):
 - 887/888 .o files were reused

These modules were not reused:
 - _main (_main.bc)

However with an alpine docker image

[centos@ip-172-28-240-100 cb]$  sudo docker run --rm -v `pwd`:/workspace -w /workspace -it cb-static-builder-amd64:latest /bin/sh
/workspace # uname -a
Linux e7b0856376dc 4.18.0-305.3.1.el8.x86_64 #1 SMP Tue Jun 1 16:14:33 UTC 2021 x86_64 Linux
[centos@ip-172-28-240-100 cb]$ crystal --version
Crystal 1.1.1 [6d9a1d583] (2021-07-26)

LLVM: 10.0.1
Default target: x86_64-unknown-linux-gnu
/workspace # shards build -s
Dependencies are satisfied
Building: cb
Parse:                             00:00:00.001736556 (   0.79MB)
Error target cb failed to compile:
Invalid memory access (signal 11) at address 0x564e649d3838
[0x564da48c8186] ???
[0x564da48c8153] ???
[0x7fa76a516304] ???

The docker file is at https://github.com/will/cb/blob/main/Dockerfile

[will]

I couldn't get much out of lldb. I'm not very good at lldb.

[centos@ip-172-28-240-100 cb]$  sudo docker run --privileged --cap-add=SYS_PTRACE --security-opt seccomp=unconfined  --rm -v `pwd`:/workspace -w /workspace -it cb-static-builder-amd64:latest /bin/sh
/workspace # apk add lldb
fetch http://dl-cdn.alpinelinux.org/alpine/v3.12/main/x86_64/APKINDEX.tar.gz
fetch http://dl-cdn.alpinelinux.org/alpine/v3.12/community/x86_64/APKINDEX.tar.gz
(1/3) Installing llvm10-libs (10.0.0-r2)
(2/3) Installing libedit (20191231.3.1-r0)
(3/3) Installing lldb (10.0.0-r0)
Executing busybox-1.31.1-r16.trigger
OK: 556 MiB in 84 packages
/workspace # lldb shards -- build
(lldb) target create "shards"
Current executable set to 'shards' (x86_64).
(lldb) settings set -- target.run-args  "build"
(lldb) run
Process 16 launched: '/usr/bin/shards' (x86_64)
Process 16 stopped
* thread #1, name = 'shards', stop reason = signal SIGSEGV: invalid address (fault address: 0x55555584df00)
    frame #0: 0x00007ffff7e95463
->  0x7ffff7e95463: movsbq (%rax), %rdi
    0x7ffff7e95467: callq  0x7ffff7e88300
    0x7ffff7e9546c: jmp    0x7ffff7e953f4
    0x7ffff7e9546e: callq  0x7ffff7e87ed0
(lldb) bt
* thread #1, name = 'shards', stop reason = signal SIGSEGV: invalid address (fault address: 0x55555584df00)
  * frame #0: 0x00007ffff7e95463
(lldb) l
(lldb) 

[straight-shoota]

We're experiencing several issues with segfaults on alpine. The test_alpine CI job is constantly failing, for example. This shouldn't happen on alpine 3.12, though (#10366).

Looks like you're using the 1.1.0 compiler from alpine edge. That build is broken: https://gitlab.alpinelinux.org/alpine/aports/-/issues/12854

[will]

Looks like you're using the 1.1.0 compiler from alpine edge

I think it's 1.1.1 but still from the --version output, but still broken. But thanks for pointing out the source of the problem.

I've gone back to 1.0.0 for just the linux builds with RUN apk add --update --no-cache --repository=http://dl-cdn.alpinelinux.org/alpine/latest-stable/community crystal shards instead, and that fixed it.

@maxfierke the dockerfile I was using was from your mtrap project, so you might also run into this, just fyi.

@straight-shoota I'd close this issue now unless you think there is some value in tracking the issue here too. If not feel free to close it.

edit: I made an account on their gitlab instance, but it won't let me subscribe for notifications :( I put in an email address.

[beta-ziliani]

Thanks @will, I'm closing it