Bilingual

The pandemic of incomplete OpenSSL error handling
22 days ago
- A bug report for APT highlighted TLS failures on FIPS systems due to MD5 errors, suggesting use of ERR_clear_error() around TLS operations.
- The author opposes this as a bad practice, arguing that discarding unrelated errors is unsafe and that the root cause should be fixed instead.
- ERR_clear_error() is widely used as a best practice in many codebases, but it leads to a systemic issue of incomplete error handling.
- Another anti-pattern involves checking top-level errors after OpenSSL operations and discarding them if deemed not severe, which also silences unrelated errors.
- Developers are encouraged to inspect code for ERR_clear_error() calls and consider using ERR_set_mark() for safer error stack management.
- A call is made to OpenSSL authors to avoid promoting practices that compromise security and trust in software.
The one-line mistake that makes bash scripts fail silently
10 days ago
- Bash scripts can fail silently, leading to undetected backup failures and other issues, especially in automated environments.
- Use 'set -euo pipefail' as a foundation to stop on errors, undefined variables, and pipe failures, but be aware of its limitations.
- Implement 'trap' for cleanup to ensure temporary files and resources are removed regardless of how the script exits.
- Always quote variables to prevent whitespace splitting and globbing, avoiding common bugs in production.
- Validate dependencies, arguments, and environment before acting to prevent failures in different setups.
- Write detailed logs with timestamps and context to aid troubleshooting, especially during off-hours incidents.
- Use file locking with 'flock' to prevent concurrent execution and data corruption in scheduled scripts.
- Utilize ShellCheck for static analysis and consider Bats for testing to catch errors before deployment.
- Conduct a final pre-production check including flags, cleanup, quoting, validation, logging, concurrency protection, and ShellCheck results.
- Share experiences and tips in communities to learn from common bash script mistakes and improve reliability.
Throw, Result, or Neither?
5 days ago
- The author argues that business failures like out-of-stock or item limit should be returned as events rather than exceptions or Result types, allowing them to be queried and used in reports or workflows.
- Exceptions are reserved for invalid operations (e.g., adding to a confirmed cart) or infrastructure errors (e.g., unavailable event store), while Result adds little value beyond success/failure classification in most popular languages.
- Middleware like rejectOn, skipOn, stopOn, and stopAfter in Emmett allows flexible handling of events without persisting them, supporting different use cases (e.g., import rejection vs. saved-list restore).
- Event sourcing enables selective persistence: the decision returns events, but the application decides whether to append them, avoiding unnecessary data retention.

error handling

The pandemic of incomplete OpenSSL error handling

The one-line mistake that makes bash scripts fail silently

Throw, Result, or Neither?