{"version":"2.1.312","anchor":"shell-command-cpu-time-telemetry-now-limited-to-linuxwsl-w","canonical_anchor":"shell-command-cpu-time-telemetry-now-limited-to-linuxwsl-w","heading":"Shell command CPU time reading limited to Linux and WSL, with stricter error handling","tier":"internal","area":"Telemetry","scope":"individual","heads_up":false,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.312\/e\/shell-command-cpu-time-telemetry-now-limited-to-linuxwsl-w","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.312","markdown":"### Shell command CPU time reading limited to Linux and WSL, with stricter error handling\n\nThe \/proc reader for shell command CPU time now runs only on Linux and WSL, takes a process limit, and stops on unexpected read errors\n\n**What**\n\nClaude Code measures how much CPU time a shell command used by reading `\/proc`, a Linux folder that describes running processes. That reader changed:\n\n- It now returns straight away unless the platform is `linux` or `wsl` (Windows Subsystem for Linux).\n\n- It now takes the process limit as a parameter.\n\n- It still skips known harmless read errors but rethrows any other read error. Before, it skipped every failure.\n\n- It logs a warning only for errors other than `EMFILE`, which means too many open files. The warning text changed.\n\n- Messages involved include \"Could not read \/proc (\" and \"Gave up reading the CPU time of a shell command's processes\".\n\n**Why**\n\nThis only affects internal measurement. Other platforms no longer try the read or log warnings about it, and unexpected read failures now stop the measurement instead of being quietly skipped.\n\n- Area: Telemetry\n- Tier: Under the hood\n- Useful: 1\/5\n- Signal: 1\/5\n- Scope: individual\n- Heads-up: no"}