F-8824: enable config files without trailing newline

[padelsbach]

No description provided.

[Copilot]

Pull request overview

Enables known_hosts parsing to handle files whose last line does not end with a trailing newline, and adds regression coverage to prevent reintroducing the issue.

Changes:

• Make load_der_file() allocate an extra byte and NUL-terminate the buffer past the file contents.
• Update ClientPublicKeyCheck() parsing to avoid clobbering the last byte and to normalize CRLF line endings.
• Add a regression test that exercises last-entry matching with \n, no trailing newline, and \r\n.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
apps/wolfssh/common.c	NUL-terminate loaded file buffers and normalize CRLF in known_hosts parsing.
tests/regress.c	Add regression test for known_hosts last-line handling (with/without trailing newline and CRLF).

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #1156

Scan targets checked:wolfssh-bugs, wolfssh-src

Findings: 5
5 finding(s) posted as inline comments (see file-level comments below)

This review was generated automatically by Fenrir. Findings are non-blocking.

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #1156

Scan targets checked:wolfssh-bugs, wolfssh-src

Findings: 5
5 finding(s) posted as inline comments (see file-level comments below)

This review was generated automatically by Fenrir. Findings are non-blocking.

[wolfSSL-Fenrir-bot]

🔵 [Low] Rejection assertion cannot distinguish a bad-key rejection from an unparsed entry · Weak or missing assertions

ClientPublicKeyCheck returns -1 both for the badMatch path and for the unknown-host path, where GetConfirmation() reads EOF from the redirected stdin and declines. The != 0 assertion therefore passes even if the last entry is never parsed, so it does not pin the behaviour its comment describes.

Fix: Capture the prompt's stdout or add a distinguishable return/observable so the bad-key rejection is asserted separately from the trust-on-first-use decline.

[wolfSSL-Fenrir-bot]

⚪ [Info] Redundant lineSz 0 guard is always true · Dead error handling

The enclosing if (line != NULL && *line) already guarantees a non-empty string, so WSTRLEN(line) is at least 1 and the lineSz > 0 sub-condition can never be false.

Related known finding #8812 (similar but distinct): Both findings identify an always-true condition caused by prior string-length/termination guarantees, but this operates on a known_hosts line in ClientPublicKeyCheck while issue 8812 operates on a path buffer in wolfSSH_GetPath. The faulting checks and immediate causes differ, and each requires a separate patch.

Fix: Drop the lineSz > 0 term and test line[lineSz - 1] == '\r' alone.

[wolfSSL-Fenrir-bot]

🔵 [Low] WriteKnownHosts dereferences a failed file handle when WOLFSSH_NO_ABORT is set · NULL pointer dereference

Assert expands to if (!(test)) Fail(...), and Fail reduces to a bare printf when WOLFSSH_NO_ABORT is defined (regress.c:50-67). If WFOPEN fails, f remains WBADFILE (NULL on POSIX) and execution falls through to WFWRITE/WFCLOSE, dereferencing it. LoadFileBuffer at regress.c:474 uses a real if guard instead.

Fix: Guard the write and close with if (f != WBADFILE) rather than relying on Assert for control flow.

[wolfSSL-Fenrir-bot]

🔵 [Low] Negative known_hosts assertion cannot distinguish a hard key-mismatch reject from a TOFU prompt reading EOF · Weak or missing assertions

ClientPublicKeyCheck returns -1 both for the intended badMatch hard reject (common.c:459) and for the unknown-host TOFU path where GetConfirmation() reads EOF from the redirected /dev/null stdin (common.c:486-491). The != 0 check passes either way, so a regression that downgrades a known-host key mismatch into a user prompt would not be caught.

Fix: Assert the exact return value and distinguish the reject path from the prompt path, e.g. by feeding a 'Y' on stdin so an unintended prompt would return success.

[wolfSSL-Fenrir-bot]

⚪ [Info] otherMatch branch newly reachable under CRLF known_hosts is not covered · Missing edge-case coverage on a function the PR also changed

The CR strip changes reachability of the otherMatch branch at common.c:440-449: under CRLF a same-key/different-host entry previously never matched, and now it does, driving the fingerprint print plus AppendKeyToFile prompt at common.c:467-469. All three fixture cases use a non-matching AAAA key for the other host, so that branch is never entered.

Fix: Add a case where the non-target host line carries the same encoded key so the otherMatch branch is exercised under CRLF and no-trailing-newline inputs.