fix: error when install path does not exist

[waveywaves]

Summary

• Add install.sh to the main repository (migrated from the now-archived chainloop-dev/docs repo at static/install.sh) with two bug fixes:
  • Early path validation: Check that the --path installation directory exists before downloading anything, with a clear error message and actionable suggestion (mkdir -p)
  • Better install error reporting: Replace silent 2>/dev/null stderr suppression in the install step with proper error handling and user-facing messages

Note: The install script was previously hosted only in the archived chainloop-dev/docs repo and served via dl.chainloop.dev. The CDN/deployment pipeline will need to be updated to serve this script from the new location.

Fixes #2977

Test plan

• Run bash install.sh --path /nonexistent/path and verify it exits early with: The installation path '/nonexistent/path' does not exist. Please create it first with: mkdir -p /nonexistent/path
• Run bash install.sh --path /tmp/test-install (after mkdir -p /tmp/test-install) and verify normal installation succeeds
• Run bash install.sh with default path (/usr/local/bin) and verify existing behavior is preserved
• Verify install failure (e.g., permission denied to a read-only directory) produces a meaningful error instead of silent failure

🤖 Generated with Claude Code

[cubic-dev-ai]

3 issues found across 1 file

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="install.sh">

<violation number="1" location="install.sh:58">
P2: `validate_checksums_file` changes CWD with `cd "$1"` and never restores it. The script later runs `rm -r ${TMP_DIR}` while still inside TMP_DIR, which fails and causes the script to exit under `set -e`, leaving the temp directory behind.</violation>

<violation number="2" location="install.sh:85">
P2: `--help` prints usage but does not exit, so the installer proceeds with downloads and installation instead of stopping.</violation>

<violation number="3" location="install.sh:92">
P2: `--path` does not validate a following argument, so with `set -u` the script exits on an unbound `$1` when `--path` is the final CLI argument.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review, or fix all with cubic.}

[waveywaves]

@jiparis This PR is ready for review — all commits GPG-signed, DCO signed-off, rebased on latest main. Fixes the install script error when the target directory doesn't exist (issue #2580).

[cubic-dev-ai]

1 issue found across 1 file (changes from recent commits).

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="install.sh">

<violation number="1" location="install.sh:57">
P2: Changing checksum validation to a subshell breaks cwd side effects and causes signature download to write into caller CWD instead of TMP_DIR.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review, or fix all with cubic.}

[cubic-dev-ai]

P2: Changing checksum validation to a subshell breaks cwd side effects and causes signature download to write into caller CWD instead of TMP_DIR.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At install.sh, line 57:

<comment>Changing checksum validation to a subshell breaks cwd side effects and causes signature download to write into caller CWD instead of TMP_DIR.</comment>

<file context>
@@ -54,7 +54,7 @@ is_command() {
 # checksums.txt file validation
 # example: check_sha256 "${TMP_DIR}" checksums.txt
-validate_checksums_file() {
+validate_checksums_file() (
   cd "$1"
   if is_command sha256sum; then
</file context>