Error output from ls -A results in error output, why?

Question

I have a script I run regularly using cron. I would like to get notified by email when these scripts fail. I do not wish to be notified every time they run and produce any output at all.

As such, I am using the script Cronic to run my jobs inside cron, which should mean only error output gets sent, and not just any output.

However, in one script I have a command like this:

if [ "$(ls -A ${local_backup_location}/nextcloud-data/)" ]; then
  # save space by removing diffs older than 6 months
  rdiff-backup --remove-older-than 6M --force ${local_backup_location}/nextcloud-data/ || echo "[$(date "+%Y-%m-%d %T")] No existing nextcloud data backup"
fi

The ls -A ${local_backup_location}/nextcloud-data/ is intended to test if a directory is empty. My problem is that this command seems to result in output which is recognized as error output cronic. Cronic defines an error as any non-trace error output or a non-zero result code. For example:

Cronic detected failure or error output for the command:
/usr/local/sbin/run_backup

RESULT CODE: 0

ERROR OUTPUT:
appdata_ocgcv9nemegb
files_external
flow.log
flow.log.1
__groupfolders
.htaccess
index.html
nextcloudadmin
nextcloud-db.bak
nextcloud.log
nextcloud.log.1
.ocdata
rdiff-backup-data
Test_User
updater.log
updater-ocgcv9nemegb ]
custom
gitea-db.sql
log ]
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed

  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
100   365    0     0  100   365      0    302  0:00:01  0:00:01 --:--:--   303
100   365    0     0  100   365      0    165  0:00:02  0:00:02 --:--:--   165
100   365    0     0  100   365      0    113  0:00:03  0:00:03 --:--:--   113
100   365    0     0  100   365      0     86  0:00:04  0:00:04 --:--:--    86
100   365    0     0  100   365      0     70  0:00:05  0:00:05 --:--:--    70
100   365    0     0  100   365      0     58  0:00:06  0:00:06 --:--:--     0
100   365    0     0  100   365      0     50  0:00:07  0:00:07 --:--:--     0
100   365    0     0  100   365      0     44  0:00:08  0:00:08 --:--:--     0
100   365    0     0  100   365      0     39  0:00:09  0:00:09 --:--:--     0
100   365    0     0  100   365      0     37  0:00:09  0:00:09 --:--:--     0
100 10.4M    0 10.4M  100   365  1016k     34  0:00:10  0:00:10 --:--:-- 2493k
100 11.6M    0 11.6M  100   365  1128k     34  0:
 00:10  0:00:10 --:--:-- 3547k

STANDARD OUTPUT:
Maintenance mode enabled
Deleting increment at time:
<snip>

So why does the command ls -A ${local_backup_location}/nextcloud-data/ produce error output in this case, and how can I prevent this? An alternative robust method to test if a directory is empty would be acceptable, but I would also like an explanation of why the command seems to produce error output.

EDIT: Adding Cronic stdout with set -ex

Some commenters have requested the actual whole script which is very long, but Cronic reports the actual stdout of the script and I use set -ex at the top of the script. The error output happens immediately after the invocation of ls -A /mnt/reos-storage-2/backups/nextcloud-data/ which is why I believe the error output to be the result of this command.

+ rdiff-backup --ssh-no-compression /var/www/nextcloud /mnt/reos-storage-2/backups/nextcloud/
+ ls -A /mnt/reos-storage-2/backups/nextcloud-data/
+ [ 67cf481e-62a3-1039-8bf2-05805d214bca
<removed>
appdata_ocgcv9nemegb
<removed>
<removed>
<removed>
<removed>
files_external
flow.log
flow.log.1
__groupfolders
.htaccess
index.html
<removed>
<removed>
nextcloudadmin
nextcloud-db.bak
nextcloud.log
nextcloud.log.1
.ocdata
<removed>
<removed>
rdiff-backup-data
<removed>
Test_User
<removed>
updater.log
updater-ocgcv9nemegb ]
+ rdiff-backup --remove-older-than 6M --force /mnt/reos-storage-2/backups/nextcloud-data/
+ date +%Y-%m-%d %T
+ echo [2021-04-21 03:23:38] Starting nextcloud data backup

Answer 1

+ [ 67cf481e-62a3-1039-8bf2-05805d214bca
<removed>
[...]
updater.log
updater-ocgcv9nemegb ]

This here, is one single command. It's split to multiple lines in the set -x/xtrace output, because the output from ls contains newlines. (Bash would print that with some quotes, Dash doesn't.)

By default, the xtrace output goes to stderr, the same as the regular error output, Cronic tries to separate them by looking at the + marker at the start of the xtrace lines. That fails here, and it thinks the files names there are part of regular error output.

What cronic does is basically this:

PATTERN="^${PS4:0:1}\\+${PS4:1}"
if grep -aq "$PATTERN" $TRACE
then
    ! grep -av "$PATTERN" $TRACE > $ERR

Disabling xtrace would be one way to fix that, but it would be a shame to do that since cronic supports it so nicely.

Instead, it might be better to use some other way to check if the directory is empty.

Keeping with ls -A, you could pipe the output to wc to count the characters there:

if [ "$(ls -A "${local_backup_location}/nextcloud-data/" | wc -c)" -gt 0 ]; then
    echo "directory not empty"
fi

or to grep:

if ls -A "${local_backup_location}/nextcloud-data/" | grep -q .; then
    echo "directory not empty";
fi

Checking if the directory is empty could be done in other ways, within the shell itself, but handling all the corner cases can be hairy. See e.g. Portable check empty directory

Answer 2

If $gitea_backup_dir doesn't exist, or makes ls unhappy in any way, ls will write an error message to STDERR.

You can discard the whole error stream by appending 2>/dev/null to your command.

IMHO, a better way is:

if ([[ -d "$gitea_backup_dir" ]] && \
    [[ $(stat --format="%h" "$gitea_backup_dir") -gt 2 ]] ) ; then
    echo "Nonempty"
else
    echo "Empty"
fi
  

First, check that the directory exists, then see if the number of hard links in the directory (the number of file and subdirectory entries in the directory) is greater than 2 (every directory has at least 2 entries: . and ..).