Implement set -e alternative

[ading2210]

xonfig

$ xonfig
+------------------+-----------------+
| xonsh            | 0.15.1          |
| Python           | 3.11.2          |
| PLY              | 3.11            |
| have readline    | True            |
| prompt toolkit   | None            |
| shell type       | readline        |
| history backend  | json            |
| pygments         | None            |
| on posix         | True            |
| on linux         | True            |
| distro           | unknown         |
| on wsl           | False           |
| on darwin        | False           |
| on windows       | False           |
| on cygwin        | False           |
| on msys2         | False           |
| is superuser     | False           |
| default encoding | utf-8           |
| xonsh encoding   | utf-8           |
| encoding errors  | surrogateescape |
| xontrib          | []              |
| RC file          | []              |
+------------------+-----------------+

Expected Behavior

In bash, doing cmd || true will ignore any errors that the command generates if set -e is on. I would expect that the behavior is similar in Xonsh with $RAISE_SUBPROC_ERROR, which is the equivalent setting.

For example:

bash

bash -c "
set -e
false || true
echo "success"
"
# success

Current Behavior

A subprocess.CalledProcessError is called if any of the commands in the expression fail.

Traceback (if applicable)

xonsh 

$RAISE_SUBPROC_ERROR = True
$XONSH_SHOW_TRACEBACK = True
false || true

# <xonsh-code>:1:0:6 - false || true
# <xonsh-code>:1:0:6 + ![false]
# <xonsh-code>:1:8:14 - false || true
# <xonsh-code>:1:8:14 + ![true]
# xonsh: To log full traceback to a file set: $XONSH_TRACEBACK_LOGFILE = <filename>
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/user/.local/pipx/venvs/xonsh/lib/python3.11/site-packages/xonsh/built_ins.py", line 206, in subproc_captured_hiddenobject
    return xonsh.procs.specs.run_subproc(cmds, captured="hiddenobject", envs=envs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/.local/pipx/venvs/xonsh/lib/python3.11/site-packages/xonsh/procs/specs.py", line 908, in run_subproc
    return _run_specs(specs, cmds)
           ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/.local/pipx/venvs/xonsh/lib/python3.11/site-packages/xonsh/procs/specs.py", line 943, in _run_specs
    command.end()
  File "/home/user/.local/pipx/venvs/xonsh/lib/python3.11/site-packages/xonsh/procs/pipelines.py", line 459, in end
    self._end(tee_output=tee_output)
  File "/home/user/.local/pipx/venvs/xonsh/lib/python3.11/site-packages/xonsh/procs/pipelines.py", line 478, in _end
    self._raise_subproc_error()
  File "/home/user/.local/pipx/venvs/xonsh/lib/python3.11/site-packages/xonsh/procs/pipelines.py", line 603, in _raise_subproc_error
    raise subprocess.CalledProcessError(rtn, spec.args, output=self.output)
subprocess.CalledProcessError: Command '['false']' returned non-zero exit status 1.

Steps to Reproduce

1. Open xonsh
2. Run $RAISE_SUBPROC_ERROR = True
3. Run false || true

For community

⬇️ Please click the 👍 reaction instead of leaving a +1 or 👍 comment

[anki-code]

Indeed, $RAISE_SUBPROC_ERROR is to raise error in any case when any command fails but set -e treat the line as whole:

Exit immediately if a pipeline (see Pipelines), which may consist of a single simple command (see Simple Commands), a list (see Lists of Commands), or a compound command (see Compound Commands) returns a non-zero status.

So we need additional setting e.g. $XONSH_RAISE_SUBPROC_PIPELINE_ERROR to do the same as bash. PR is welcome!

[anki-code]

JFYI. For this:

set -e
ls nonono || true     # false || true
echo "success"

the xonsh equvalent for now is this:

$RAISE_SUBPROC_ERROR = True
try:
    ls nonono
except:               # OR except subprocess.CalledProcessError:
    pass
echo "success"

[anki-code]

You can create syntax sugar for this using macro call:

import subprocess
def ignore_error(cmd):
    """Ignore failing of the command."""
    try:
        execx(cmd)
    except subprocess.CalledProcessError:
        pass
        
        
$RAISE_SUBPROC_ERROR = False      
ignore_error!(echo 1 and (ls nonono or echo 2))
echo success
# success
        
$RAISE_SUBPROC_ERROR = True        
ignore_error!(echo 1 and (ls nonono or echo 2))
echo success
# success

[ading2210]

I'm aware that I can do a try-except block to ignore errors, though that still doesn't help with the underlying problem of boolean logic being broken with subprocess return codes.

[anki-code]

I figured this out the case.

What bash doing. In fact it implicitly overrides the execution of a logical expression and the execution of process. If we run echo 1 and ls nonono the result of this logical will be treated as "return code" and if it's not 0 and set -e the error will be raised.

In xonsh we have separation between process running and logical expression running. If we run echo 1 and ls nonono we have 2 separate processes and logical expression and RAISE_SUBPROC_ERROR will work on process level.

In fact in this issue was requested an additional mode like RAISE_COMMAND_ERROR that will work for the whole logical expression.

Current behavior:

$RAISE_SUBPROC_ERROR=False
echo 1 and (ls no or echo 2)
# run logical: `echo 1`, `ls no`, `echo 2`.
__xonsh__.history[-1]
# rtn=0
echo 1 and (echo 2 or ls no)
# run logical: `echo 1`, `echo 2`.
__xonsh__.history[-1]
# rtn=0

$RAISE_SUBPROC_ERROR=True
echo 1 and (ls no or echo 2)
# run logical: `echo 1`, `ls no` (raise and stop).
__xonsh__.history[-1]
# rtn=2
echo 1 and (echo 2 or ls no)
# run logical: `echo 1`, `echo 2`.
__xonsh__.history[-1]
# rtn=0