Running lit commands in docker build segfault at exit.

Question

Running lit commands in docker build segfault at exit.

creationix opened this issue 3 years ago · 3 comments

I reproduced this with latest luvi (v2.12.0) on both alpine and ubuntu docker environments with both building from source and using the prebuilt get-lit luvi binaries. I don't think it's a luvi issue entirely.

A simple docker based luvi+lit app looks something like this:

FROM creationix/lit
ADD package.lua .
RUN lit install
ADD main.lua .
EXPOSE 8080
CMD ["luvi", "."]

This normally works great, but with the latest version of lit, it will fail after lit install finishes:

$ docker build webserver/
Sending build context to Docker daemon  4.096kB
Step 1/6 : FROM creationix/lit
 ---> 59a1da203be1
Step 2/6 : ADD package.lua .
 ---> e22d625ce600
Step 3/6 : RUN lit install
 ---> Running in f58a663faf85
lit version: 3.8.5
luvi version: v2.12.0
command: install
create config: /root/.litconfig
connecting: wss://lit.luvit.io/
fetching: 1 object
...
fetching: 1 object
including dependency: coro-channel (3.0.3)
...
including dependency: weblit-server (3.1.3)
installing package: creationix/weblit-app@v3.2.1
...
installing package: luvit/http-codec@v3.0.6
done: success

The command '/bin/sh -c lit install' returned a non-zero code: 139

I'm wondering if it's related to some recent changes to how stdout is managed in pretty-print? Not sure really.

I can work around the issue in the Dockerfile by ignoring the return value like RUN lit install || true. As far as I can tell, the result is correct.

I don't think it's related to a lack of tty. I can use docker run creationix/lit lit ... with or without a pseudo terminal and there are no segfaults for any lit command. All lit commands segfault while exiting during the docker build step.

Answer 1 · 2021-07-30T16:48:18.000Z

Not sure if it's any help, but I recently faced a similar issue when testing electron apps. It would segfault only on the Ubuntu VM, apparently because the output could not be rendered. My workaround was something like export DISPLAY=:45 && xvfb-run --server-num 45 before running the command in question, though I'm not a Linux expert so that might be bad.

I used GitHub actions and they run inside a Docker environment, so I figured it might be at least tangentially related... Could also be something else, since I also had the same issue when accessing the VM via ssh.

Answer 2 · 2021-08-06T05:26:15.000Z

Run into the same issue trying to setup lua in a docker container. Here are some artifacts from me digging into the issue.

FROM ubuntu:latest
RUN apt-get update && apt-get upgrade -y
RUN apt-get install -y tzdata && \
    rm /etc/localtime && \
    ln -fs /usr/share/zoneinfo/US/Eastern /etc/localtime && \
    dpkg-reconfigure -f noninteractive tzdata
RUN apt-get install -y lua5.3 liblua5.3-dev sqlite3 libsqlite3-dev vim curl build-essential unzip gdb
RUN cd /root && curl -L https://github.com/luvit/lit/raw/master/get-lit.sh | sh && \
    mv ./lit /usr/local/bin/lit && \
    mv ./luvi /usr/local/bin/luvi && \
    mv ./luvit /usr/local/bin/luvit
RUN gdb -batch -ex "run" -ex "bt" --args lit help 2>&1 || true

Build output:

Step 8/8 : RUN gdb -batch -ex "run" -ex "bt" --args lit help 2>&1 || true
 ---> Running in f7c497d4134e
warning: Error disabling address space randomization: Operation not permitted
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
lit version: 3.8.5
luvi version: v2.11.0-21-g5d1052f
command: help

... help message

done: success


Program received signal SIGABRT, Aborted.
__GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
50      ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.
#0  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
#1  0x00007fec82031859 in __GI_abort () at abort.c:79
#2  0x000000000045d1bd in uv.io_poll.cold ()
#3  0x000000000052c83a in uv_run ()
#4  0x0000000000494f7e in luv_run ()
#5  0x000000000052462a in lj_BC_FUNCC ()
#6  0x000000000052561a in lj_ff_coroutine_wrap_aux ()
#7  0x000000000051023f in lua_pcall ()
#8  0x000000000045d36b in main ()
Removing intermediate container f7c497d4134e
 ---> f967125d3a6f

Answer 3 · 2021-08-06T07:48:33.000Z

Some more output when CMAKE_BUILD_TYPE=Debug for luvi.

FROM ubuntu:latest
RUN apt-get update && apt-get upgrade -y
RUN apt-get install -y tzdata && \
    rm /etc/localtime && \
    ln -fs /usr/share/zoneinfo/US/Eastern /etc/localtime && \
    dpkg-reconfigure -f noninteractive tzdata
RUN apt-get install -y lua5.3 liblua5.3-dev sqlite3 libsqlite3-dev vim curl build-essential unzip gdb cmake
RUN apt-get install -y git zlib1g-dev libbz2-dev libeditline-dev

# Manual build of luvi
RUN cd /root && git clone --recursive --branch v2.12.0 https://github.com/luvit/luvi.git && \
    cd luvi && \
    sed -i 's/DCMAKE_BUILD_TYPE=Release/DCMAKE_BUILD_TYPE=Debug/' Makefile && \
    make regular && \
    make && \
    make test
RUN cd /root/ && \
    curl -L -f -o lit.zip "https://lit.luvit.io/packages/luvit/lit/v3.8.5.zip" && \
    ./luvi/build/luvi lit.zip -- make lit.zip || true
RUN cd /root && \
    gdb -batch -ex "set follow-fork-mode child" -ex "run" -ex "bt" -ex "frame 2" -ex "list"  --args /root/lit help 2>&1 || true

Program received signal SIGABRT, Aborted.
__GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
50      ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.
#0  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
#1  0x00007ff87df84859 in __GI_abort () at abort.c:79
#2  0x0000561b59bd84a2 in uv__io_poll (loop=0x7ff87df4ed98, timeout=0) at /root/luvi/deps/luv/deps/libuv/src/unix/linux-core.c:254
#3  0x0000561b59bbf63c in uv_run (loop=0x7ff87df4ed98, mode=UV_RUN_DEFAULT) at /root/luvi/deps/luv/deps/libuv/src/unix/core.c:385
#4  0x0000561b59af0242 in luv_run (L=0x7ff87df571f0) at /root/luvi/deps/luv/src/loop.c:36
#5  0x0000561b59bb4f9a in lj_BC_FUNCC () at buildvm_x86.dasc:851
#6  0x0000561b59bb5f8a in lj_ff_coroutine_wrap_aux () at buildvm_x86.dasc:1808
#7  0x0000561b59b27648 in lua_pcall (L=0x7ff87df3d380, nargs=1, nresults=1, errfunc=2) at /root/luvi/deps/luv/deps/luajit/src/lj_api.c:1169
#8  0x0000561b59ae2a90 in main (argc=2, argv=0x561b5b5ff2a0) at /root/luvi/src/main.c:191
#2  0x0000561b59bd84a2 in uv__io_poll (loop=0x7ff87df4ed98, timeout=0) at /root/luvi/deps/luv/deps/libuv/src/unix/linux-core.c:254
254             abort();
249         /* XXX Future optimization: do EPOLL_CTL_MOD lazily if we stop watching
250          * events, skip the syscall and squelch the events after epoll_wait().
251          */
252         if (epoll_ctl(loop->backend_fd, op, w->fd, &e)) {
253           if (errno != EEXIST)
254             abort();
255
256           assert(op == EPOLL_CTL_ADD);
257
258           /* We've reactivated a file descriptor that's been watched before. */

The actual location in libuv repository get moved by this commit libuv/libuv@caf22dd