How to Limit CPU Usage on Linux Processes Using CPULimit

·1 views

CPULimit is a Linux utility that allows users to cap the CPU consumption of specific processes by targeting their process ID (PID). It can be installed on Debian-based systems via apt and on RHEL or CentOS systems via yum, with the EPEL repository required for the latter. Once installed, users run the tool with flags specifying the target PID and a percentage ceiling, for example restricting a process to 40% CPU usage. The utility also supports a background mode using the --background or -b flag, allowing it to run without occupying the terminal. Tools like htop or top can be used to verify that the CPU limit is being enforced after cpulimit is applied.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Three silent macOS traps that break Claude Code hook scripts explained

A developer debugging Claude Code hook scripts on macOS discovered three distinct failure modes that all share the trait of failing without any error output. The first trap involves zsh's read-only reserved variable 'status', which silently ignores assignment attempts unlike in bash, causing condition checks to never trigger. The second issue arises when Claude Code is launched as a GUI app, as child processes inherit only a minimal PATH that excludes tools like Node.js installed via nvm or Homebrew. The third trap is the absence of a 'timeout' command on macOS, which differs from Linux environments where it is available by default. The author recommends renaming reserved variables, setting a comprehensive 'env.PATH' in Claude Code's settings.json, and accounting for macOS-specific command availability when writing hook scripts.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why CFOs Need a Zero-Trust Gateway Before Deploying AI in Accounts Payable

Companies exploring AI-driven accounts payable and invoice processing face a critical risk: large language models occasionally generate incorrect math totals, and even a 1% error rate can translate into millions in financial liability. A proposed hybrid architecture addresses this by separating AI-driven data extraction from deterministic validation, ensuring no AI agent directly interacts with a corporate ledger or ERP system. A Node.js gateway layer sits between the AI and the database, re-calculating subtotals, tax figures, and line items to the exact cent before any transaction is executed. An automated circuit breaker halts processing immediately if a significant variance or potential fraud signal is detected, while a SHA-256 cryptographic chain locks each verified transaction into an immutable audit log. The approach aims to make AI agents a controlled operational asset rather than a compliance risk, giving finance and audit teams full visibility into every automated decision.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Refines iOS Fretboard App Using AI Pair Programming and Swift Testing

A developer shared their Week 16 progress log detailing advances on a personal iOS fretboard practice application built with SwiftUI. Working alongside OpenAI's Codex, they defined and implemented the completion logic for a Note Practice mode, including progress display, feedback timers, fret-range controls, and automatic question transitions. When Codex generated unit tests using the older XCTest framework, the developer revised them to use Apple's newer Swift Testing framework, updating project documentation to preserve the policy for future AI sessions. Key technical work included implementing NotePracticeCalculator, introducing FretCellVisualState for explicit UI state representation, and resolving a Swift Concurrency warning by marking FretPosition as nonisolated. The developer also made incremental progress on TryHackMe's AI Security Learning Path alongside the app work.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Pentester Builds Custom Traffic Listener After Burp Suite Fails on Unstable VPN App

A security researcher encountered persistent proxy failures while pentesting an internal financial web app for a student loan servicer, with Burp Suite, Caido, and other tools all failing to capture traffic reliably. The app exhibited an unusual pattern where requests would succeed briefly, then time out for up to an hour, and any modified request — such as one containing a quote or script tag — would receive no response at all. After three largely unproductive days, the researcher concluded that the unstable connection was being worsened by a man-in-the-middle proxy decrypting and resending traffic. Instead of intercepting requests, they built a lightweight browser extension that captures outgoing traffic and immediately forwards it to a local listener program, avoiding the performance issues of storing data in localStorage. The solution resolved both the CORS preflight problem and browser slowdowns by keeping the extension stateless and offloading all storage to the local listener.

0 comments Read more at DEV Community