I have encountered a bug when using entr paired with nano/pico on a Raspberry Pi Zero 1.3: Not only when saving a file, but also when opening a file with nano or pico entr will start executing its command.

I have tested this on 2 different Raspberry Pi Zero 1.3 Models. Here this bug occurs. (I have tested Version 4.1 and 4.9)

Additionally, I have tested this on my macOS machine (entr 4.8) as well as a Ubuntu Server (entr 4.4) where the bug does not occur. This is how I concluded that is a specific bug on this Raspberry Pi Zero version of Linux.

In this case, I'm using entr to restart a Ruby process when source files change.

Problem: entr does not restart the process when the Ruby process crashes (it's a development environment after all).

My environment is minikube in MacOS. Live reloading works using the entr inode workaround.

Is the problem in my configuration or what could cause this?

Entrypoint:

#! /usr/bin/env bash

set -euo pipefail

_term() {
 >&2 echo "TERM"
 exit 0
}

trap "_term" TERM

OPERATOR="${1:-}"
COMMAND="bin/operators/${OPERATOR}"

if [[ -z "${OPERATOR}" ]]; then
  echo "Error: Missing parameter operator name."
  echo "Usage: $0 operator_name"
  exit 1
elif [[ ! -x "${COMMAND}" ]]; then
  echo "Error: Command '${COMMAND}' is not executable."
  exit 1
fi

if [[ "${ENV}" = "development" ]]; then
  export ENTR_INOTIFY_WORKAROUND="yespls"
  (
    while true; do
      set +e
        find . | entr -rd "${COMMAND}"
      set -e
    done
  ) &
else
  set +e
    "${COMMAND}" &
  set -e
fi

wait $!

When building this project using ./configure && make test I get the error from the title. To reproduce with docker and Debian 11:

build_oci_img() {
  docker build "$@" - <<EOF
  FROM debian:11
  RUN DEBIAN_FRONTEND=noninteractive apt-get \
      -o DPkg::options::="--force-confdef" \
      -o DPkg::options::="--force-confold" \
      update -y && \
    DEBIAN_FRONTEND=noninteractive apt-get \
        -o DPkg::options::="--force-confdef" \
        -o DPkg::options::="--force-confold" \
        install --no-install-recommends -y \
        make git ca-certificates gcc build-essential
  RUN git clone https://github.com/eradman/entr/
  WORKDIR entr
  RUN ./configure && make test
EOF
}

build_oci_img

docker run --rm --interactive --tty "$(build_oci_img -q)"

Output:

# Docker build output before

Step 5/5 : RUN ./configure && make test
 ---> Running in 1ef231d4aeb7
cp Makefile.linux Makefile
cc  -D_GNU_SOURCE -D_LINUX_PORT -Imissing -DRELEASE=\"5.3\"  missing/strlcpy.c missing/kqueue_inotify.c entr.c -o entr
./system_test.sh
./system_test.sh: 19: Syntax error: "}" unexpected
make: *** [Makefile.bsd:9: test] Error 2
The command '/bin/sh -c ./configure && make test' returned a non-zero code: 2

Would be nice it would running on rpi4 64bit. Tried:

find v-hub.ko | entr sh -c 'clear' [1] 15504 done find v-hub.ko | 15505 segmentation fault entr sh -c 'clear'

checked on:

Linux pi 5.10.4-v8+ #1389 SMP PREEMPT Wed Jan 6 13:52:18 GMT 2021 aarch64 GNU/Linux

I'm trying to monitor some files with entr:

do_it(){ echo Eita!; }
while true; do ls folder/* more-folder/* | entr -pd do_it; done
>> entr: exec do_it: No such file or directory

However, this works:

while true; do ls folder1/* folder2/* | entr -pd echo Eita!; done

What am I doing wrong?

I am trying to use entr to help me with epub editing. So I this function will trigger an update command to a epub viwer as well as refresh a window in a web browser. (I'd love to see more real examples that would help me doing it...)

I'd like to use entr to run a command that is interactive. If I don't say 'yes' to the command though, it holds the terminal and so it doesn't re-run unless I interact with that pane. It would be nice to be able to have entr kill the command on updates but still allow me to say 'yes' when I'm ready to let the command apply changes.

(Unless of course this feature already exists, in which case, great!)

I'd like to be able to use entr on Docker images without bash/sh, which means that piping won't be available to me; ideally I would be able to specify files to watch as an arg, e.g.

["entr", "--files=a.txt,b.txt,c.txt", "-rz", "my_util.sh"]

(A smaller point in favor of this feature is that while piping makes perfect sense for a call like ls | entr ..., if i just want to watch a single file, it looks a little silly to say echo one-file.txt | entr ...)

I often find myself using entr to write helper scripts that work on files. I would love it if I could write those scripts to read from stdin and write to stdout, but there isn't a really easy way to do that with entr since it would receive shell redirection if it was done as an argument to the cli itself. It would be nice if there were a built in entr flag to tell it to send stdin to a command from a specific file to make this use case easier.

I get segmentation faults with the latest version of entr (4.7 or git checkout). This happens on a MacBook M1 (Arm chip), running macOS Big Sur, not much changed from the default configuration.

Backtrace:

(lldb) process launch -i input.txt  -- -s 'make'
Process 36108 launched: '/Users/jer/code/entr/entr' (arm64)
cur: 9223372036854775807
max: 9223372036854775807
Process 36108 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
    frame #0: 0x00000001000023d8 entr`process_input(file=0x00000001f9273240, files=0x0000000000000000, max_files=9223372036854775807) at entr.c:305:19
   302 				continue;
   303 			}
   304 			if (S_ISREG(sb.st_mode) != 0) {
-> 305 				files[n_files] = malloc(sizeof(WatchFile));
   306 				strlcpy(files[n_files]->fn, buf, MEMBER_SIZE(WatchFile, fn));
   307 				files[n_files]->is_dir = 0;
   308 				files[n_files]->file_count = 0;

I added some logging:

n_files here is 0. max_files is -1.

After getrlimit(RLIMIT_NOFILE) I get:

rl.rlim_cur = 9223372036854775807
rl.rlim_max = 9223372036854775807

that. ~~That's UINT64_MAX~~ That's INT64_MAX. sysconf(_SC_OPEN_MAX) also returns that.

As rlim_cur is the minimum out of all of that it's INT64_MAX before being passed to process_input. Now process_input takes an int, so it's coerced and becomes -1.

Seems my file descriptors are unlimited by default:

$ ulimit -n
unlimited

Changing it to some more reasonable low number makes entr work as expected:

$ ulimit -n 1024
$

Question:

• Should entr enforce a more reasonable limit for this case?
• Should process_file maybe not use a different integer type than the one passed in?

The patch file version is http://eradman.com/entrproject/patches/entr-3.9-docker while the entr is at 4.x

Currently it does apply and the binary seems to work, see https://github.com/matti/docker-entr/blob/master/build/ubuntu/Dockerfile#L13-L17

Also, could it be possible to include the patch in entr and have the behaviour available as an option? I think I've wasted something like an 2 hours trying to figure out what was wrong with docker-for-mac and entr - just found out https://github.com/eradman/entr/issues/3 by accident

I try to set a build watcher for tflint and get hit by 1024 limit.

$ fd "\.go$" | entr go build -o ./flt
entr: Too many files listed; the hard limit for your login class is 1024. Please consult
http://eradman.com/entrproject/limits.html

But my limit is 8192.

$ cat /proc/sys/fs/inotify/max_user_watches  
8192

entr does not use inotify. Why?