Edit me

QuickySitter has three layers of diagnostics, in increasing order of how invasive they are:

  1. The project-wide Out(level, …) verbose ladder: runtime logging that creators can dial up from the AVpos notecard without editing scripts.
  2. The [QS]debug script: an owner-only /88 chat tool that inspects Linkset Data and can fire synthetic load at the sitter chain.
  3. Per-script bDebug-style developer toggles: compile-time flags a contributor flips while chasing a specific bug.

The verbose ladder: Out(level, msg)

Every fork script that emits runtime chatter routes it through a small helper:

// Verbose convention: 0=error/warn floor (default), 1=boot banner,
// 2=runtime status, 3=debug. OutForce() bypasses for critical messages.
integer verbose = 0;
Out(integer level, string msg)
{
    if (verbose >= level)
        llOwnerSay(llGetScriptName() + "[" + version + "] " + msg);
}
OutForce(string msg)   // unconditional, for must-see errors
{
    llOwnerSay(llGetScriptName() + "[" + version + "] " + msg);
}

The level scale is the same across the fork:

Level Meaning Examples
0 Error / warning floor: always shown. boot “notecard missing” ERROR, offset “emergency shrink” WARN.
1 Boot / ready banner: once per script start. [QS]sitA[…] Ready, Mem=…, offset Ready. LSD room=….
2 Runtime status: per-operation chatter. Loading… during a pose/expression load.
3 Debug: fine-grained internal state. reserved for deep tracing.

Setting the level from the notecard

The single source is the AVpos notecard directive VERBOSE n. [QS]boot parses it during seed and writes it to the qs:cfg:verbose LSD key (a singleton, not per-channel). Almost every fork script reads that key in state_entry to initialise its own verbose global before its first Out() call (the exception is [QS]root-control, which ships with verbose = 1 hardcoded and never reads the key):

// in state_entry, before any Out(...)
string v = llLinksetDataRead("qs:cfg:verbose");
if (v != "") verbose = (integer)v;

Default is VERBOSE 0, silent except for the level-0 error/warning floor. Add VERBOSE 1 to the AVpos notecard to see ready banners, VERBOSE 2 for runtime status, VERBOSE 3 for everything. Because the value lives in LSD, changing it affects every script at once and survives a script reset (boot won’t clobber a user-chosen level on the skip-seed path).

Why a ladder instead of bare llOwnerSay

Passive llOwnerSay calls multiply by N on furniture-heavy regions: each sitter slot’s [QS]sitA would spam the owner every time a pose plays. Gating every line behind Out(level, …) keeps shipping builds quiet at the default floor and lets diagnostics turn on temporarily, region-wide, with one notecard token. Don’t add bare llOwnerSay calls outside the ladder; route through Out() (or OutForce() for a genuine must-see error) so the verbosity stays under one knob. See AVpos Reference for the VERBOSE token.

[QS]debug: the /88 LSD inspector

[QS]debug.lsl is an owner-only chat tool, not a link-message responder. Drop it into a QuickySitter prim and it listens on chat channel /88 filtered to the owner; all output goes back to the owner via llOwnerSay. It does not receive or answer diagnostic link-messages, and other scripts never call into it. It is safe to leave in a prim (harmless if unused) and removes cleanly.

Type /88 help for the menu. Commands:

Command What it does
keys [pattern] List qs:* keys, or keys matching a regex.
count <ch> Pose count for one channel.
meta <ch> Show qs:meta:<ch> (the boot marker).
cfg <ch> Show qs:cfg:<ch> (the 17-slot config blob).
sitter <ch> Show qs:sitter:<ch>.
pose <ch> <i> Show qs:p:<ch>:<i> with the fields parsed out (name / type / anim / pos / rot).
poses <ch> Dump every pose row for a channel.
grep <text> Find all qs:p:* rows whose value contains <text>.
raw <key> Show the raw value of any LSD key.
mem LSD bytes used / free, with a little usage bar.
delch <ch> Delete all qs:*:<ch> keys for one channel (also drops qs:boot:asset so the next reset re-seeds).
nuke / nuke yes Wipe all Linkset Data on the object (confirmation required).
stress {start\|stop\|status\|speed} Synthetic-load generator (see below).

This replaces the older idea of a central log-routing responder: inspection is pull-based (you ask /88), so there’s no debug channel for other scripts to publish to. The polished, notecard-formatted equivalent of a bulk dump is still the [DUMP] button. See Adjustment Workflow.

[QS]debug stress test

/88 stress makes [QS]debug impersonate up to 7 fake sitters and fire synthetic sitter traffic at the rest of the linkset, to load-test hudproxy and the downstream [QS]sitA / [QS]sitB / [QS]offset chain. Fake UUIDs come from llGenerateKey, so each run exercises a fresh QSO:<short>:* keyspace in [QS]offset.

Subcommand Effect
stress start [n] Spawn n fake sitters (default 6, max 7) and begin.
stress stop Run the cleanup phase (one un-sit per tick).
stress status Current phase, sitter/op counts, free LSD, debug script memory.
stress speed <ms> Chaos-phase tick interval (min 100 ms).

It runs in three timer-driven phases:

  • Ramp: one fake sitter joins per tick: 90060 (sit) + 90070 (slot assign).
  • Chaos: random ops per tick: ~70 % pose changes (90055), some quiet swaps (90031, the quiet variant so it doesn’t also reopen pose menus), and the rest 90262 saves with non-zero offsets (so the QSO:* keyspace grows rather than hitting the ZERO/ZERO delete sentinel).
  • Cleanup: one un-sit per tick (90065) until empty.

Caveat: if [QS]debug is reset mid-stress, the fake sitters orphan in hudproxy’s state and consume listen slots, so reset hudproxy (or the prim) to recover. See LinkMessage Numbers for what each of those numbers means.

Inspecting LSD by hand

If you don’t have [QS]debug in the prim, the same idea from a temporary script or debug HUD:

default
{
    touch_start(integer n)
    {
        // Dump all qs:* keys (slow, don't do this in shipping code)
        list keys = llLinksetDataFindKeys("^qs:", 0, 100);
        integer i;
        for (i = 0; i < llGetListLength(keys); ++i)
        {
            string k = llList2String(keys, i);
            llOwnerSay(k + " = " + llLinksetDataRead(k));
        }
    }
}

llLinksetDataFindKeys takes a regex over LSD keys. Use the qs: prefix as the anchor to scope the dump.

Developer toggles (bDebug)

Before the verbose ladder, individual scripts used a binary integer bDebug = FALSE; guarding debugSay(...) calls. Most scripts have since migrated to Out(): for example [QS]offset replaced its bDebug/debugSay scheme outright. It now emits exactly two ladder lines:

  • Ready banner with current LSD headroom: Out(1, "Ready. LSD room=…").
  • Emergency shrinks (memory pressure forcing RAM eviction before a save): Out(0, "WARN: …"), so they show even at the default floor.

Useful when debugging “the save didn’t persist” reports: set VERBOSE 1 to see the ready banner’s LSD headroom, and watch for the emergency-shrink WARN (which shows at the default floor regardless).

A bare bDebug-style compile-time flag is still a fine pattern when you’re adding throwaway tracing to one script while chasing a bug, as long as it doesn’t ship enabled. Conventions:

  • Ship quiet. A developer flag defaults to FALSE; the notecard VERBOSE level is the supported user-facing knob.
  • Don’t add bare llOwnerSay outside Out(). Once a passive log lands it tends to stay and shows up in user chat forever.
  • Don’t gate on runtime conditions like getNumberOfSitters() > 1. Keep the toggle binary and obvious in the source.

When investigating a bug:

  1. Add VERBOSE 1 (or 2/3) to the AVpos notecard and reset the prim, which lights up Out() across every script at once.
  2. For state questions, drop in [QS]debug and poke at LSD with /88 keys, /88 poses <ch>, /88 cfg <ch>, /88 mem.
  3. To reproduce load/offset bugs, /88 stress start, watch, then /88 stress stop.
  4. Remove the VERBOSE token (or set it to 0) and any throwaway tracing before committing. Review catches a stray passive log, but it’s easier to do up front.

See also