User feedback: column widths shouldn't reserve space for label variants
the renderer never uses. Add v4-style short_name to mem and let the
auto-size adapt:
mem.available.short_name = 'avail' (9 -> 5 chars)
mem.inactive.short_name = 'inactiv' (8 -> 7 chars)
mem.buffers.short_name = 'buffer' (7 -> 6 chars)
The mem render_curses_v5 now resolves every label via
field_label(..., prefer_short=True), so adding/changing a short_name
in the schema is enough — no renderer edit needed.
The label-column floors in cpu and mem renderers
(_CPU_LABEL_MIN_WIDTHS / _MEM_LABEL_MIN_WIDTHS) are gone too: labels
are constant strings so the content-driven auto-size is naturally
stable. Only value columns still need a fixed floor.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Previously a prominent careful level promoted the plugin title to
CAREFUL color. Per user feedback, careful is too noisy for a title
escalation: a freshly booted machine often sits in careful for one or
two cycles. Now the title stays white (HEADER) until any prominent
field reaches warning or critical.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Plugin titles (MEM/CPU/LOAD) now adapt their color to the worst
prominent alert level in the payload:
- No prominent escalation → bold white (HEADER role, v4 TITLE parity).
- Worst prominent level is careful/warning/critical → bold + that color.
Non-prominent escalations do not promote the title color (only watched +
prominent fields count).
Implementation:
- ColorRole.HEADER curses color: cyan → white (matches v4 TITLE).
- Cell gains a 'bold' flag, decoupling the bold attribute from the
HEADER role so alert-colored titles can still render bold.
- New helpers in curses_renderer_v5: _max_prominent_level(payload),
title_role(payload). Handle scalar and nested (collection) _levels
shapes.
- cpu/mem/load renderers compute the title cell color via title_role()
and set bold=True explicitly.
- All alignment helpers (generic + cpu + mem) preserve the bold flag
when rebuilding cells.
Tests cover: title_role priority ordering, prominent-only filtering,
collection level shape, painter A_BOLD applies on explicit bold flag.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2-col x 4-row grid layout:
MEM 53.2% active 5.8G
total 15.3G inactive 4.4G
avail 7.2G buffers 185M
free 2.6G cached 4.2G
Variants:
- Line 3 col 1: 'avail' (when payload has 'available' — Linux/macOS) or
'used' (BSD/other) as v4 does.
Fixed column widths (value=6 chars, label cols floored at v4 sizes) so
the block doesn't jiggle cycle-to-cycle.
The percent value cell carries the watched/prominent flag from the
schema, so the painter applies A_REVERSE under any non-DEFAULT alert
level.
Reference: glances/plugins/mem/__init__.py::msg_curse.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
