diff --git a/README.md b/README.md
index d761bda..d75183a 100644
--- a/README.md
+++ b/README.md
@@ -76,7 +76,7 @@ Output will be like this:
 ### What this library provides
 
 #### Macros to run external commands
-- run_cmd! --> CmdResult
+- [run_cmd!](https://docs.rs/cmd_lib/latest/cmd_lib/macro.run_cmd.html) -> [CmdResult](https://docs.rs/cmd_lib/latest/cmd_lib/type.CmdResult.html)
 
 ```rust
 let msg = "I love rust";
@@ -101,7 +101,7 @@ run_cmd! {
 }?;
 ```
 
-- run_fun! --> FunResult
+- [run_fun!](https://docs.rs/cmd_lib/latest/cmd_lib/macro.run_fun.html) -> [FunResult](https://docs.rs/cmd_lib/latest/cmd_lib/type.FunResult.html)
 
 ```rust
 let version = run_fun!(rustc --version)?;
@@ -209,8 +209,9 @@ Ignore errors for command execution.
 
 ##### echo
 Print messages to stdout.
-
+```console
 -n     do not output the trailing newline
+```
 
 ##### error, warn, info, debug, trace
 
@@ -227,7 +228,7 @@ run_cmd!(info "This is an infomation message")?;
 ```
 
 #### Macros to register your own commands
-Declare your function with `#[export_cmd(..)]` attribute, and import it with `use_custom_cmd!` macro:
+Declare your function with `#[export_cmd(..)]` attribute, and import it with [`use_custom_cmd!`] macro:
 
 ```rust
 #[export_cmd(my_cmd)]
@@ -244,13 +245,15 @@ println!("get result: {}", run_fun!(my_cmd)?);
 
 #### Low-level process spawning macros
 
-`spawn!` macro executes the whole command as a child process, returning a handle to it. By
+[`spawn!`] macro executes the whole command as a child process, returning a handle to it. By
 default, stdin, stdout and stderr are inherited from the parent. The process will run in the
-background, so you can run other stuff concurrently. You can call `wait()` to wait
+background, so you can run other stuff concurrently. You can call [`wait()`](`CmdChildren::wait()`) to wait
 for the process to finish.
 
-With `spawn_with_output!` you can get output by calling `wait_with_output()`, or even do stream
-processing with `wait_with_pipe()`.
+With [`spawn_with_output!`] you can get output by calling [`wait_with_output()`](`FunChildren::wait_with_output()`), or even do stream
+processing with [`wait_with_pipe()`](`FunChildren::wait_with_pipe()`).
+
+There are also other useful APIs, and you can check the docs for more details.
 
 ```rust
 let mut proc = spawn!(ping -c 10 192.168.0.1)?;
@@ -275,9 +278,9 @@ spawn_with_output!(journalctl)?.wait_with_pipe(&mut |pipe| {
 
 
 #### Macros to define, get and set thread-local global variables
-- `tls_init!` to define thread local global variable
-- `tls_get!` to get the value
-- `tls_set!` to set the value
+- [`tls_init!`] to define thread local global variable
+- [`tls_get!`] to get the value
+- [`tls_set!`] to set the value
 ```rust
 tls_init!(DELAY, f64, 1.0);
 const DELAY_FACTOR: f64 = 0.8;
@@ -322,7 +325,7 @@ You can use the [glob](https://github.com/rust-lang-nursery/glob) package instea
 
 This library tries very hard to not set global states, so parallel `cargo test` can be executed just fine.
 The only known APIs not supported in multi-thread environment are the
-`tls_init/tls_get/tls_set` macros, and you should only use them for *thread local* variables.
+[`tls_init`]/[`tls_get`]/[`tls_set`] macros, and you should only use them for *thread local* variables.
 
 
 License: MIT OR Apache-2.0
diff --git a/macros/src/lib.rs b/macros/src/lib.rs
index 60ce06b..cb96ed9 100644
--- a/macros/src/lib.rs
+++ b/macros/src/lib.rs
@@ -2,7 +2,7 @@ use proc_macro2::{Span, TokenStream, TokenTree};
 use proc_macro_error::{abort, proc_macro_error};
 use quote::{quote, ToTokens};
 
-/// Mark main function to log error result by default
+/// Mark main function to log error result by default.
 ///
 /// ```
 /// # use cmd_lib::*;
@@ -40,7 +40,7 @@ pub fn main(
     .into()
 }
 
-/// Export the function as an command to be run by `run_cmd!` or `run_fun!`
+/// Export the function as an command to be run by [`run_cmd!`] or [`run_fun!`].
 ///
 /// ```
 /// # use cmd_lib::*;
@@ -79,7 +79,7 @@ pub fn export_cmd(
     new_functions.into()
 }
 
-/// Import user registered custom command
+/// Import user registered custom command.
 /// ```
 /// # use cmd_lib::*;
 /// #[export_cmd(my_cmd)]
@@ -118,7 +118,7 @@ pub fn use_custom_cmd(item: proc_macro::TokenStream) -> proc_macro::TokenStream
     .into()
 }
 
-/// Run commands, returning result handle to check status
+/// Run commands, returning [`CmdResult`](../cmd_lib/type.CmdResult.html) to check status.
 /// ```
 /// # use cmd_lib::run_cmd;
 /// let msg = "I love rust";
@@ -155,7 +155,7 @@ pub fn run_cmd(input: proc_macro::TokenStream) -> proc_macro::TokenStream {
     .into()
 }
 
-/// Run commands, returning result handle to capture output and to check status
+/// Run commands, returning [`FunResult`](../cmd_lib/type.FunResult.html) to capture output and to check status.
 /// ```
 /// # use cmd_lib::run_fun;
 /// let version = run_fun!(rustc --version)?;
@@ -177,8 +177,7 @@ pub fn run_fun(input: proc_macro::TokenStream) -> proc_macro::TokenStream {
     .into()
 }
 
-/// Run commands with/without pipes as a child process, returning a handle to check the final
-/// result
+/// Run commands with/without pipes as a child process, returning [`CmdChildren`](../cmd_lib/struct.CmdChildren.html) result.
 /// ```
 /// # use cmd_lib::*;
 ///
@@ -199,8 +198,7 @@ pub fn spawn(input: proc_macro::TokenStream) -> proc_macro::TokenStream {
     .into()
 }
 
-/// Run commands with/without pipes as a child process, returning a handle to capture the
-/// final output
+/// Run commands with/without pipes as a child process, returning [`FunChildren`](../cmd_lib/struct.FunChildren.html) result.
 /// ```
 /// # use cmd_lib::*;
 /// let mut procs = vec![];
@@ -231,7 +229,7 @@ pub fn spawn_with_output(input: proc_macro::TokenStream) -> proc_macro::TokenStr
 
 #[proc_macro]
 #[proc_macro_error]
-/// Log a fatal message at the error level, and exit process
+/// Log a fatal message at the error level, and exit process.
 ///
 /// e.g:
 /// ```
diff --git a/src/child.rs b/src/child.rs
index c615e6f..d186c74 100644
--- a/src/child.rs
+++ b/src/child.rs
@@ -1,4 +1,4 @@
-use crate::{info, warn};
+use crate::info;
 use crate::{process, CmdResult, FunResult};
 use os_pipe::PipeReader;
 use std::io::{BufRead, BufReader, Error, ErrorKind, Read, Result};
@@ -8,7 +8,7 @@ use std::thread::JoinHandle;
 /// Representation of running or exited children processes, connected with pipes
 /// optionally.
 ///
-/// Calling `spawn!` macro will return `Result<CmdChildren>`
+/// Calling [`spawn!`](../cmd_lib/macro.spawn.html) macro will return `Result<CmdChildren>`
 pub struct CmdChildren {
     children: Vec<CmdChild>,
     ignore_error: bool,
@@ -70,7 +70,7 @@ impl CmdChildren {
 /// Representation of running or exited children processes with output, connected with pipes
 /// optionally.
 ///
-/// Calling `spawn_with_output!` macro will return `Result<FunChildren>`
+/// Calling [spawn_with_output!](../cmd_lib/macro.spawn_with_output.html) macro will return `Result<FunChildren>`
 pub struct FunChildren {
     children: Vec<CmdChild>,
     ignore_error: bool,
@@ -88,17 +88,13 @@ impl FunChildren {
                 Err(e)
             }
             Ok(output) => {
-                let mut s = String::from_utf8_lossy(&output).to_string();
-                if s.ends_with('\n') {
-                    s.pop();
-                }
                 let ret = CmdChildren::wait_children(&mut self.children);
                 if let Err(e) = ret {
                     if !self.ignore_error {
                         return Err(e);
                     }
                 }
-                Ok(s)
+                Ok(output)
             }
         }
     }
@@ -107,7 +103,7 @@ impl FunChildren {
     /// provided function.
     pub fn wait_with_pipe(&mut self, f: &mut dyn FnMut(Box<dyn Read>)) -> CmdResult {
         let child = self.children.pop().unwrap();
-        let polling_stderr = StderrLogging::new(&child.cmd, child.stderr);
+        let stderr_thread = StderrThread::new(&child.cmd, child.stderr, false);
         match child.handle {
             CmdChildHandle::Proc(mut proc) => {
                 if let Some(stdout) = child.stdout {
@@ -126,10 +122,20 @@ impl FunChildren {
                 }
             }
         };
-        drop(polling_stderr);
+        drop(stderr_thread);
         CmdChildren::wait_children(&mut self.children)
     }
 
+    /// Waits for the children processes to exit completely, returning the command result, stdout
+    /// read result and stderr read result.
+    pub fn wait_with_all(&mut self) -> (CmdResult, FunResult, FunResult) {
+        // wait for the last child result
+        let handle = self.children.pop().unwrap();
+        let wait_all = handle.wait_with_all(true);
+        let _ = CmdChildren::wait_children(&mut self.children);
+        wait_all
+    }
+
     /// Returns the OS-assigned process identifiers associated with these children processes
     pub fn pids(&self) -> Vec<u32> {
         self.children.iter().filter_map(|x| x.pid()).collect()
@@ -158,8 +164,9 @@ impl CmdChild {
         }
     }
 
-    fn wait(self, is_last: bool) -> CmdResult {
-        let res = self.handle.wait_with_stderr(self.stderr, &self.cmd);
+    fn wait(mut self, is_last: bool) -> CmdResult {
+        let _stderr_thread = StderrThread::new(&self.cmd, self.stderr.take(), false);
+        let res = self.handle.wait(&self.cmd);
         if let Err(e) = res {
             if is_last || process::pipefail_enabled() {
                 return Err(e);
@@ -168,27 +175,35 @@ impl CmdChild {
         Ok(())
     }
 
-    fn wait_with_output(self, ignore_error: bool) -> Result<Vec<u8>> {
-        let buf = {
-            if let Some(mut out) = self.stdout {
-                let mut buf = vec![];
-                if let Err(e) = out.read_to_end(&mut buf) {
-                    if !ignore_error {
-                        return Err(process::new_cmd_io_error(&e, &self.cmd));
+    fn wait_with_output(self, ignore_error: bool) -> FunResult {
+        let (res, stdout, _) = self.wait_with_all(false);
+        if !ignore_error {
+            res?;
+        }
+        stdout
+    }
+
+    fn wait_with_all(mut self, capture: bool) -> (CmdResult, FunResult, FunResult) {
+        let mut stderr_thread = StderrThread::new(&self.cmd, self.stderr.take(), capture);
+        let stdout_output = {
+            if let Some(mut out) = self.stdout.take() {
+                let mut s = String::new();
+                match out.read_to_string(&mut s) {
+                    Err(e) => Err(e),
+                    Ok(_) => {
+                        if s.ends_with('\n') {
+                            s.pop();
+                        }
+                        Ok(s)
                     }
                 }
-                buf
             } else {
-                vec![]
+                Ok("".into())
             }
         };
-        let res = self.handle.wait_with_stderr(self.stderr, &self.cmd);
-        if let Err(e) = res {
-            if !ignore_error {
-                return Err(e);
-            }
-        }
-        Ok(buf)
+        let stderr_output = stderr_thread.join();
+        let res = self.handle.wait(&self.cmd);
+        (res, stdout_output, stderr_output)
     }
 
     fn kill(self) -> CmdResult {
@@ -207,8 +222,7 @@ pub(crate) enum CmdChildHandle {
 }
 
 impl CmdChildHandle {
-    fn wait_with_stderr(self, stderr: Option<PipeReader>, cmd: &str) -> CmdResult {
-        let polling_stderr = StderrLogging::new(cmd, stderr);
+    fn wait(self, cmd: &str) -> CmdResult {
         match self {
             CmdChildHandle::Proc(mut proc) => {
                 let status = proc.wait();
@@ -242,7 +256,6 @@ impl CmdChildHandle {
             }
             CmdChildHandle::SyncFn => {}
         }
-        drop(polling_stderr);
         Ok(())
     }
 
@@ -272,19 +285,31 @@ impl CmdChildHandle {
     }
 }
 
-struct StderrLogging {
-    thread: Option<JoinHandle<()>>,
+struct StderrThread {
+    thread: Option<JoinHandle<String>>,
     cmd: String,
 }
 
-impl StderrLogging {
-    fn new(cmd: &str, stderr: Option<PipeReader>) -> Self {
+impl StderrThread {
+    fn new(cmd: &str, stderr: Option<PipeReader>, capture: bool) -> Self {
         if let Some(stderr) = stderr {
             let thread = std::thread::spawn(move || {
+                let mut output = String::new();
                 BufReader::new(stderr)
                     .lines()
                     .map_while(Result::ok)
-                    .for_each(|line| info!("{}", line))
+                    .for_each(|line| {
+                        if !capture {
+                            info!("{line}");
+                        } else {
+                            output.push_str(&line);
+                            output.push('\n');
+                        }
+                    });
+                if output.ends_with('\n') {
+                    output.pop();
+                }
+                output
             });
             Self {
                 cmd: cmd.into(),
@@ -297,14 +322,28 @@ impl StderrLogging {
             }
         }
     }
-}
 
-impl Drop for StderrLogging {
-    fn drop(&mut self) {
+    fn join(&mut self) -> FunResult {
         if let Some(thread) = self.thread.take() {
-            if let Err(e) = thread.join() {
-                warn!("[{}] logging thread exited with error: {:?}", self.cmd, e);
+            match thread.join() {
+                Err(e) => {
+                    return Err(Error::new(
+                        ErrorKind::Other,
+                        format!(
+                            "Running [{}] stderr thread joined with error: {e:?}",
+                            self.cmd
+                        ),
+                    ))
+                }
+                Ok(output) => return Ok(output),
             }
         }
+        Ok("".into())
+    }
+}
+
+impl Drop for StderrThread {
+    fn drop(&mut self) {
+        let _ = self.join();
     }
 }
diff --git a/src/lib.rs b/src/lib.rs
index f396d90..a8ba0e2 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -83,7 +83,7 @@
 //! ## What this library provides
 //!
 //! ### Macros to run external commands
-//! - run_cmd! --> CmdResult
+//! - [run_cmd!](https://docs.rs/cmd_lib/latest/cmd_lib/macro.run_cmd.html) -> [CmdResult](https://docs.rs/cmd_lib/latest/cmd_lib/type.CmdResult.html)
 //!
 //! ```no_run
 //! # use cmd_lib::run_cmd;
@@ -110,7 +110,7 @@
 //! # Ok::<(), std::io::Error>(())
 //! ```
 //!
-//! - run_fun! --> FunResult
+//! - [run_fun!](https://docs.rs/cmd_lib/latest/cmd_lib/macro.run_fun.html) -> [FunResult](https://docs.rs/cmd_lib/latest/cmd_lib/type.FunResult.html)
 //!
 //! ```
 //! # use cmd_lib::run_fun;
@@ -230,8 +230,9 @@
 //!
 //! #### echo
 //! Print messages to stdout.
-//!
+//! ```console
 //! -n     do not output the trailing newline
+//! ```
 //!
 //! #### error, warn, info, debug, trace
 //!
@@ -250,7 +251,7 @@
 //! ```
 //!
 //! ### Macros to register your own commands
-//! Declare your function with `#[export_cmd(..)]` attribute, and import it with `use_custom_cmd!` macro:
+//! Declare your function with `#[export_cmd(..)]` attribute, and import it with [`use_custom_cmd!`] macro:
 //!
 //! ```
 //! # use cmd_lib::*;
@@ -270,13 +271,15 @@
 //!
 //! ### Low-level process spawning macros
 //!
-//! `spawn!` macro executes the whole command as a child process, returning a handle to it. By
+//! [`spawn!`] macro executes the whole command as a child process, returning a handle to it. By
 //! default, stdin, stdout and stderr are inherited from the parent. The process will run in the
-//! background, so you can run other stuff concurrently. You can call `wait()` to wait
+//! background, so you can run other stuff concurrently. You can call [`wait()`](`CmdChildren::wait()`) to wait
 //! for the process to finish.
 //!
-//! With `spawn_with_output!` you can get output by calling `wait_with_output()`, or even do stream
-//! processing with `wait_with_pipe()`.
+//! With [`spawn_with_output!`] you can get output by calling [`wait_with_output()`](`FunChildren::wait_with_output()`), or even do stream
+//! processing with [`wait_with_pipe()`](`FunChildren::wait_with_pipe()`).
+//!
+//! There are also other useful APIs, and you can check the docs for more details.
 //!
 //! ```no_run
 //! # use cmd_lib::*;
@@ -304,9 +307,9 @@
 //!
 //!
 //! ### Macros to define, get and set thread-local global variables
-//! - `tls_init!` to define thread local global variable
-//! - `tls_get!` to get the value
-//! - `tls_set!` to set the value
+//! - [`tls_init!`] to define thread local global variable
+//! - [`tls_get!`] to get the value
+//! - [`tls_set!`] to set the value
 //! ```
 //! # use cmd_lib::{ tls_init, tls_get, tls_set };
 //! tls_init!(DELAY, f64, 1.0);
@@ -356,15 +359,15 @@
 //!
 //! This library tries very hard to not set global states, so parallel `cargo test` can be executed just fine.
 //! The only known APIs not supported in multi-thread environment are the
-//! `tls_init/tls_get/tls_set` macros, and you should only use them for *thread local* variables.
+//! [`tls_init`]/[`tls_get`]/[`tls_set`] macros, and you should only use them for *thread local* variables.
 //!
 
 pub use cmd_lib_macros::{
     cmd_die, export_cmd, main, run_cmd, run_fun, spawn, spawn_with_output, use_custom_cmd,
 };
-/// Return type for run_fun!() macro
+/// Return type for [`run_fun!()`] macro.
 pub type FunResult = std::io::Result<String>;
-/// Return type for run_cmd!() macro
+/// Return type for [`run_cmd!()`] macro.
 pub type CmdResult = std::io::Result<()>;
 pub use child::{CmdChildren, FunChildren};
 #[doc(hidden)]
diff --git a/src/process.rs b/src/process.rs
index 9128f06..6a9e304 100644
--- a/src/process.rs
+++ b/src/process.rs
@@ -19,7 +19,7 @@ use std::thread;
 const CD_CMD: &str = "cd";
 const IGNORE_CMD: &str = "ignore";
 
-/// Environment for builtin or custom commands
+/// Environment for builtin or custom commands.
 pub struct CmdEnv {
     stdin: CmdIn,
     stdout: CmdOut,
@@ -29,32 +29,32 @@ pub struct CmdEnv {
     current_dir: PathBuf,
 }
 impl CmdEnv {
-    /// Returns the arguments for this command
+    /// Returns the arguments for this command.
     pub fn args(&self) -> &[String] {
         &self.args
     }
 
-    /// Fetches the environment variable key for this command
+    /// Fetches the environment variable key for this command.
     pub fn var(&self, key: &str) -> Option<&String> {
         self.vars.get(key)
     }
 
-    /// Returns the current working directory for this command
+    /// Returns the current working directory for this command.
     pub fn current_dir(&self) -> &Path {
         &self.current_dir
     }
 
-    /// Returns a new handle to the standard input for this command
+    /// Returns a new handle to the standard input for this command.
     pub fn stdin(&mut self) -> impl Read + '_ {
         &mut self.stdin
     }
 
-    /// Returns a new handle to the standard output for this command
+    /// Returns a new handle to the standard output for this command.
     pub fn stdout(&mut self) -> impl Write + '_ {
         &mut self.stdout
     }
 
-    /// Returns a new handle to the standard error for this command
+    /// Returns a new handle to the standard error for this command.
     pub fn stderr(&mut self) -> impl Write + '_ {
         &mut self.stderr
     }
@@ -83,14 +83,14 @@ pub fn export_cmd(cmd: &'static str, func: FnFun) {
     CMD_MAP.lock().unwrap().insert(OsString::from(cmd), func);
 }
 
-/// Set debug mode or not, false by default
+/// Set debug mode or not, false by default.
 ///
 /// Setting environment variable CMD_LIB_DEBUG=0|1 has the same effect
 pub fn set_debug(enable: bool) {
     std::env::set_var("CMD_LIB_DEBUG", if enable { "1" } else { "0" });
 }
 
-/// Set pipefail or not, true by default
+/// Set pipefail or not, true by default.
 ///
 /// Setting environment variable CMD_LIB_PIPEFAIL=0|1 has the same effect
 pub fn set_pipefail(enable: bool) {
diff --git a/src/thread_local.rs b/src/thread_local.rs
index 3869f5c..95f121d 100644
--- a/src/thread_local.rs
+++ b/src/thread_local.rs
@@ -1,4 +1,4 @@
-/// Declare a new thread local storage variable
+/// Declare a new thread local storage variable.
 /// ```
 /// # use cmd_lib::*;
 /// use std::collections::HashMap;
@@ -15,7 +15,7 @@ macro_rules! tls_init {
     };
 }
 
-/// Get the value of a thread local storage variable
+/// Get the value of a thread local storage variable.
 ///
 /// ```
 /// # use cmd_lib::*;
@@ -35,7 +35,7 @@ macro_rules! tls_get {
     };
 }
 
-/// Set the value of a thread local storage variable
+/// Set the value of a thread local storage variable.
 /// ```
 /// # use cmd_lib::*;
 /// # let changes = "";