e.g.

[cmd wurzel(
|
[command]

i would have expected to see 127 or somesuch.

In order to cover cases where the exact output needs to be preserved without FUDI parsing, a binary output would be helpful.

When the output of the left outlet (exit code) is used to trigger a new 'exec' message, [command] prints an error:

old process still running

Desired behaviour:

• The left output should only fire after it is safe to execute another command.

The problem can be mitigated by inserting a [delay 0].

6eed98d introduced a new problem to the send method. Its output looks like it is using uninitialized memory.

How to reproduce.

1. run nc -l -n -p 8000 in a terminal
2. send exec nc localhost 8000 to [command]
3. send send one two three four five to [command]

What I get in the terminal:

��[l�one two three four jn5Vfive

What I would expect:

one two three four five

Depending on used audio backend, reliability is better or worse. Definitely much worse with JACK.

for whatever reason it uses post(...) rather than pd_error(x,...).

i think it is the responsibility of the user to prevent such a case (e.g. by implementing their own queue).
in order to find the instance that lacks such a command queue, it would be nice to be able to just click the error message.

Commands like jack_lsp use leading tabs to distinguish between ports and what ports they are connected to.

On Linux, [command] finds executables in /usr/bin/ and /usr/local/bin. On macOS 10.15 (Catalina), [command] only finds executables in /usr/bin, but not in /usr/local/bin. exec <command> where <command> is some binary from /usr/local/bin returns 2 on the left outlet, while the other outlets are silent.

This little patch segfaults Pd:

[exec false(
|
[command]

consider a patch like this:

[exec echo foo(
|
[command]
|        |       |
|        |
|        [print stdout]
|
[print err]

this will have the printout:

stdout: foo
err: 0

compare this to a patch like this:

[exec echo 123(
|
[command]
|        |       |
|        |
|        [print stdout]
|
[print err]

this will have the printout:

err: 123
err: 0

i would have expected this instead:

stdout: 123
err: 0

Quoting an idea from @umlaeute:

"as an aside: it might be interesting to be able to set per-process environment variables (so you can manually add /usr/local/bin and /Users/reduzent/.backdoor/bin/ to your PATH).

setting envvars could be done via a simple [env( message to [command]. the only missing part is to query the current value of an envvar (my gut feeling tells me that [command] is not the right object for this; however you could already do something like [exec echo $PATH()"

Yeah, querying can already be done with [exec env( or [exec echo $VAR(

It might be desirable in some situations to run [command] in blocking mode, so that order of execution is maintained, even if there is a [command] in the chain. Maybe this could be done with a flag -d (depth-first?) .

[loadbang]
|
[t b b]
|     |
|     [exec date (
|     |
|     [command  -d]
|  /
[print]

Would yield:

print: Do 24 Nov 2022 11:10:24 CET
print: bang

it seems that there is a memory leak in

it's probably not a big issue, as the data is allocated in the fork()ed process, which gets destroyed soon enough, but still...

also, 255 characters per argument seem to be a bit low. is there any specific reason to not use MAXPDSTRING?

command forks of the command to execute, so Pd can continue to do it's thing while the command runs. nice.

however, i think this needs to be documented more explicitly.

so far the only thing in the help-patch that deals with this topic is a single sentence saying:

long running commands are not blocking

i think the "long" is misleading, as the blocking behavior is the same for short-running and long-running commands.
so probably "running commands are not blocking Pd" would be better.

also, the documentation should make it explicit that the output of [command] will break order of execution.

the following will always output the return code of the previous command (so on first run it will output 0)

|
[t b b]
|     [exec false(
|     [command]
|int  ]
|

Escape characters should be removed before symbols are passed to command.