"kill -9 -$pid" timeout cleanup masks the real TimeoutException when no PID is found

[AurelDemiri]

Deployer Version

v7.5.12 (bug code is also present unchanged in master / v8)

Target OS

target OS is irrelevant

Which PHP version are you using?

PHP 8.5

Content of deploy.php or deploy.yaml

<?php
namespace Deployer;

host('example')
    ->set('hostname', 'example.com')
    ->set('remote_user', 'deploy')
    ->set('deploy_path', '~/site');

// Any remote command that exceeds the timeout AND whose process is no
// longer matched by `ps x | grep $shellId` when Deployer attempts cleanup
// will surface this bug. A real-world trigger I hit was
// `deploy:writable` running `chown -L` over a shared/symlinked storage
// tree on shared hosting — `chown` was slow enough to exceed
// default_timeout, and by the time cleanup ran the shell had exited so
// the PID lookup returned empty.
set('default_timeout', 60);

task('repro', function () {
    run('sleep 120 && exit');
});

Steps to reproduce

1. Configure any host with default_timeout lower than the duration of
   the next run() call.

2. Run a task whose remote command will exceed that timeout. In my case
   it was the built-in deploy:writable task running chown -L over a
   storage shared-symlink tree on Combell, taking >300s.

3. Deployer's SSH client (master: src/Ssh/SshClient.php:77-86; v7:
   src/Component/Ssh/Client.php:96-107) catches the
   ProcessTimedOutException and runs:

   $pid = $this->run($host, "ps x | grep $shellId | grep -v grep | awk '{print $1}'", ...);
   $this->run($host, "kill -9 -$pid", ...);

4. If the ps … awk pipeline returns an empty string — which can happen
   because the remote shell already exited, ps x is restricted, or the
   timed-out command no longer matches the shell id — then $pid is
   "" and the next call literally runs kill -9 -. Bash rejects that
   with kill: -: arguments must be process or job IDs.

5. $this->run(...) sees the non-zero exit and throws RunException,
   which replaces the TimeoutException that the surrounding code was
   about to throw.

Observed output (real example from a deploy:writable run):

[production] bash: line 1: kill: -: arguments must be process or job IDs
[production] error in writable.php on line 91:
[production] exit code 1 (General error)
done deploy:writable 5m 1s

Expected: a TimeoutException clearly stating the command exceeded N
seconds.
Actual: a cryptic kill: -: error attached to whatever task happened to
be running, with no mention of the timeout. The exact 5m 1s duration
matching default_timeout: 300 is the only clue.

This bug was already reported as issue #3782 (auto-closed via the stale
bot; later "me too" replies from sebastiaanluca, svycka and Joon-99 went
unanswered). Reposting here per the new bugs-in-Discussions policy.

[antonmedv]

Fixed in 00072f9

Please try https://github.com/deployphp/deployer/releases/tag/v8.0.5