Skip to content

loadbearer v1.5.2

Choose a tag to compare

@github-actions github-actions released this 10 Sep 19:01
· 46 commits to main since this release

Fixes only, and no measurement changes at all — 1.5.1 scores and result files
are directly comparable. Both fixes are about something that was on screen or
on disk and shouldn't have been.

  • The diagnostic log no longer records machine identifiers. At
    --log-level debug it wrote the SMBIOS UUID, the chassis serial, the asset
    tag, the OS install ID and the hostname into loadbearer.log, while
    PRIVACY.md promised in the same breath that the log held no personal data
    beyond its own file path in the header. It records whether each of those was
    readable now, rather than what it said. That is the diagnostic that was
    wanted anyway, and the distinction that matters is which artefact leaks: a
    result file is one you choose to share and can strip, a log is the one a bug
    report asks you to attach. The exception is an identifier rejected as an OEM
    placeholder ("To Be Filled By O.E.M." and friends), now logged as-is —
    it describes the firmware rather than the machine, and telling "the firmware
    reported nothing" from "it reported something useless" is worth being able
    to do.
  • The progress gauge's label stays readable once the bar fills past it.
    ratatui swaps the gauge style's foreground and background for the label
    cells over the filled part, so the label reads against the fill instead of
    disappearing into it — but only when both halves of the pair are set, and
    only the foreground was. The swap therefore drew the label in the terminal's
    own default foreground on top of Color::Cyan, whose actual colour is
    whatever the user's scheme says: under Dracula that is 1.30:1, and the
    elapsed/eta text vanished. It measures 11.4:1 there now, 6.1:1 on Windows
    Terminal's Campbell, 4.7:1 on Gruvbox and 4.1:1 on Solarized Dark. Both
    gauges had it, the run screen and the soak screen; it was reported as a
    Linux problem, but Campbell's 1.99:1 only made it look survivable.
  • cut-release.sh checks for a terminal before the ten minutes of work
    rather than after, since its confirmations read /dev/tty and can't be
    answered without one.
  • The website now has a section on
    loadbearer-fleet, which
    reports across an estate of collected results.

Full Changelog: v1.5.1...v1.5.2