Bash Configuration

This dotfiles repository includes several bash-related configurations and tools.

Bash Startup Files

The bash configuration is organized as follows:

~/.bashrc: Main bash configuration file
~/.bashrc.d/*.sh: Modular bash configuration files
~/.profile: Login shell profile

Non-Interactive Short-Circuit

~/.bashrc aborts on its first line for non-interactive shells, so none of ~/.bashrc.d/*.sh runs.

Interactive vs non-interactive

These describe whether the shell reads commands from a prompt, independently of whether it sources ~/.bashrc:

Interactive — a prompt you type at ($- contains i). E.g. ssh host with no command.
Non-interactive — handed a one-off command and exits (no i). E.g. ssh host '<cmd>', or the remote shells that rsync/scp spawn.

The guard keys off this property ($-), not off whether ~/.bashrc was sourced.

The sshd abnormality

Normally a non-interactive bash does not read ~/.bashrc at all. The exception: when bash detects it was started by sshd it sources ~/.bashrceven though it is non-interactive (a compile-time feature enabled in practically every build). So rsync/scp/ssh host '<cmd>' are non-interactive — line editing is off, so bind warns — yet they abnormally run ~/.bashrc. Any stdout written during that sourcing corrupts the program's stream; Atuin's init emitting a bind warning broke rsync:

text

$ rsync -av some_file.txt remote-dotfiles-host:/home/user/dir
.bashrc.d/30-atuin.sh: line 747: bind: warning: line editing not enabled
protocol version mismatch -- is your shell clean?

The guard keeps the stream clean and makes these one-off shells start instantly.

Limitation

Skipping ~/.bashrc.d/ means tooling activated there (mise, nix, homebrew) is unavailable to non-interactive remote commands, so ssh host 'mise-managed-tool …' won't find those tools. Anything that must survive into non-interactive sessions belongs in ~/.profile / ~/.profile.d/*.sh; interactive-only setup stays in ~/.bashrc.d/.

This is an acceptable trade-off because the common non-interactive path (rsync/scp) is pure file transfer and needs no tooling. If the limitation ever becomes a blocker — e.g. you routinely run ssh host '<managed-tool> …' — drop the top-of-file guard and instead gate only the offenders per-script (the bind calls in 30-atuin.sh, the bare where at the end of ~/.bashrc). Tooling then loads non-interactively, at the cost of full startup on every remote command and ongoing vigilance against new stdout offenders.

chezmoi run scripts: re-establish PATH yourself

The same short-circuit hits chezmoi's run_* scripts under home/.chezmoiscripts/: they execute as non-interactive shells and never source ~/.bashrc.d/, so tools installed by the bootstrap (mise, brew, nix) are not on PATH just because the interactive config would put them there. A script that calls such a tool without first putting it on PATH dies with command not found — even though the tool is installed.

Each run script must therefore re-establish PATH itself, at the top, before invoking the tool. To keep this uniform and in one place, the setup lives in shared .chezmoitemplates partials, included with {{ template "<name>" . }}:

mise-shellenv — prepends ~/.local/bin (the mise entry binary) and the mise shims dir. Include it in any script that calls mise directly, or a mise-managed tool directly (e.g. pitchfork). Note mise run <task> already exposes that task's managed tools, so for mise run … only the mise binary itself needs to be reachable.
nix-profile — sources the OS-correct nix profile (macOS multi-user daemon vs Linux single-user), guarded so it is a no-op when nix is absent (safe under set -e).
brew-shellenv — evals brew shellenv for the OS/arch-correct prefix.

The interactive ~/.bashrc.d/ files (10-nix, 17-mise, 08-homebrew) use the same partials, so interactive and non-interactive setup share one source of truth. After installing a new bootstrap tool, add a partial for it and include it in both the relevant run script and its ~/.bashrc.d/ file.

Bash Startup Profiling

This feature allows you to measure the startup time of individual files sourced during bash initialization to identify performance bottlenecks.

Usage

Enable profiling by setting the environment variable:
bash
```
export BASH_PROFILE_TIMING=1
```
Start a new bash session to generate the timing log:
bash
```
bash -l
```
Process the timing data using the mise task:
bash
```
mise run bash:profile
```

Output

The task processes the timing data in a pipeline and displays a formatted summary with rankings and percentages. No intermediate files are created.

Example Output

After enabling profiling and running the commands, the output might look like this:

Without activation hooks caching:

bash

$ BASH_PROFILE_TIMING=1 exec bash
Bash profiling enabled. Log will be written to ~/.local/share/dotfiles/bash/profile_timing.log
host  : example-host
user  : user
shell : bash 5.3.3(1)-release
wd    : /home/user

: mise run bash:profile
[bash:profile] $ ~/.config/mise/tasks/bash/profile
Processing bash startup timing log...

=== Bash Startup Timing Summary ===
Rank  Sourced File                                        Time (ms)   Relative   Cumulative
----- -----------------------------------------           ---------   --------  -----------
1     /Users/user/.bashrc.d/17-mise.sh                      627.363     51.65%       51.65%
2     /Users/user/.bashrc.d/15-devbox.sh                    337.218     27.76%       79.41%
3     /Users/user/.bashrc.d/27-chezmoi.sh                    66.921      5.51%       84.92%
4     /Users/user/.bashrc.d/82-carapace.sh                   42.791      3.52%       88.44%
5     /Users/user/.bashrc.d/19-pixi.sh                       42.516      3.50%       91.94%
6     /Users/user/.bashrc.d/40-homebrew.sh                   25.505      2.10%       94.04%
7     /Users/user/.bashrc.d/90-direnv.sh                     20.107      1.66%       95.70%
8     /Users/user/.bashrc.d/30-atuin.sh                      19.015      1.57%       97.26%
9     /Users/user/.bashrc.d/75-zellij.sh                     16.325      1.34%       98.60%
10    /Users/user/.bashrc.d/25-zoxide.sh                      7.302      0.60%       99.21%
11    /Users/user/.bashrc.d/18-usage.sh                       4.817      0.40%       99.60%
12    /Users/user/.bashrc.d/32-nvim.sh                        0.328      0.03%       99.63%
13    /Users/user/.bashrc.d/35-functions.sh                   0.309      0.03%       99.65%
14    /Users/user/.bashrc.d/05-cache.sh                       0.272      0.02%       99.68%
15    /Users/user/.bashrc.d/20-aliases.sh                     0.197      0.02%       99.69%
16    /Users/user/.bashrc.d/60-tmux.sh                        0.180      0.01%       99.71%
17    /Users/user/.bashrc.d/10-nix.sh                         0.170      0.01%       99.72%
18    /Users/user/.bashrc.d/47-amazon-q.sh                    0.168      0.01%       99.74%
19    /Users/user/.bashrc.d/45-julia.sh                       0.164      0.01%       99.75%
20    /Users/user/.bashrc.d/55-task.sh                        0.139      0.01%       99.76%
21    /Users/user/.bashrc.d/48-amazon-q_cargo.sh              0.134      0.01%       99.77%
22    /Users/user/.bashrc.d/50-aichat.sh                      0.109      0.01%       99.78%
23    /Users/user/.bashrc.d/57-yazi.sh                        0.063      0.01%       99.79%
24    /Users/user/.bashrc.d/52-jupyter.sh                     0.044      0.00%       99.79%
      TOTAL SOURCED                                        1212.157     99.79%       99.79%
      TOTAL                                                1214.708       100%         100%

With caching (60% faster):

bash

=== Bash Startup Timing Summary ===
Rank  Sourced File                                        Time (ms)   Relative   Cumulative
----- -----------------------------------------           ---------   --------  -----------
1     /Users/user/.bashrc.d/17-mise.sh                      101.585     21.02%       21.02%
2     /Users/user/.bashrc.d/40-homebrew.sh                   58.143     12.03%       33.05%
3     /Users/user/.bashrc.d/15-devbox.sh                     50.943     10.54%       43.59%
4     /Users/user/.bashrc.d/30-atuin.sh                      47.790      9.89%       53.48%
5     /Users/user/.bashrc.d/90-direnv.sh                     39.156      8.10%       61.58%
6     /Users/user/.bashrc.d/25-zoxide.sh                     38.471      7.96%       69.54%
7     /Users/user/.bashrc.d/18-usage.sh                      37.715      7.80%       77.34%
8     /Users/user/.bashrc.d/75-zellij.sh                     37.555      7.77%       85.11%
9     /Users/user/.bashrc.d/82-carapace.sh                   35.852      7.42%       92.53%
10    /Users/user/.bashrc.d/19-pixi.sh                       18.856      3.90%       96.43%
11    /Users/user/.bashrc.d/27-chezmoi.sh                    14.085      2.91%       99.35%
12    /Users/user/.bashrc.d/35-functions.sh                   0.098      0.02%       99.37%
13    /Users/user/.bashrc.d/05-cache.sh                       0.068      0.01%       99.38%
14    /Users/user/.bashrc.d/55-task.sh                        0.067      0.01%       99.40%
15    /Users/user/.bashrc.d/45-julia.sh                       0.053      0.01%       99.41%
16    /Users/user/.bashrc.d/20-aliases.sh                     0.048      0.01%       99.42%
17    /Users/user/.bashrc.d/10-nix.sh                         0.032      0.01%       99.42%
18    /Users/user/.bashrc.d/57-yazi.sh                        0.030      0.01%       99.43%
19    /Users/user/.bashrc.d/60-tmux.sh                        0.030      0.01%       99.44%
20    /Users/user/.bashrc.d/47-amazon-q.sh                    0.028      0.01%       99.44%
21    /Users/user/.bashrc.d/32-nvim.sh                        0.025      0.01%       99.45%
22    /Users/user/.bashrc.d/50-aichat.sh                      0.021      0.00%       99.45%
23    /Users/user/.bashrc.d/48-amazon-q_cargo.sh              0.020      0.00%       99.46%
24    /Users/user/.bashrc.d/52-jupyter.sh                     0.017      0.00%       99.46%
      TOTAL SOURCED                                         480.688     99.46%       99.46%
      TOTAL                                                 483.298       100%         100%

Notes

Profiling only works with Bash 5+ (requires $EPOCHREALTIME)
The timing log is overwritten at the start of each profiling session
Log is stored at ~/.local/share/dotfiles/bash/profile_timing.log
Only files in ~/.bashrc.d/*.sh and /etc/bash_completion are profiled
Disable profiling by unsetting the variable: unset BASH_PROFILE_TIMING
Caveat: Atuin files bypass profiling to preserve execution context. Atuin's bash-preexec hooks are extremely sensitive to execution context - even function wrappers change the call stack enough to break command recording

Activation Hook Caching

To improve startup performance, mise activation hooks are cached. The cache is automatically refreshed when:

The mise binary (~/.local/bin/mise) is updated
Tool installations change (~/.local/share/mise/installs directory timestamp)
Project configuration changes (mise.toml in current directory)

The caching implementation includes robust error handling for edge cases:

First execution: When mise isn't installed yet, activation is skipped gracefully
Command failures: If mise activate bash fails, an empty cache is created to prevent errors
Missing output: Commands that produce no output are handled safely
Binary availability: Mise activation only occurs if the binary is available

Limitations: The cache may not refresh for all configuration changes (e.g., global config modifications, environment-specific configs). If activation hooks seem stale, manually clear the cache:

bash

mise run bash:clean

This removes all cached activation hooks and forces regeneration on next shell startup.