deploy.

2026-09-16 00:00:49 +08:00
parent 1a75503105
commit bfa3106a69
60 changed files with 696 additions and 268 deletions
@@ -258,11 +258,21 @@ extended table (P1–P48). A selected row nobody has entered reports <code>Coord
 spelling of <code>G20</code> and reports <code>Unit--InchNotSupported</code> rather than silently switching units.
 <code>G05.1 Q1</code> / <code>Q0</code> is AICC / Nano Smoothing (recognized). Bare <code>G05 P{n}</code> is a different feature
 (HPCC) and is consumed without changing programmed coordinates, reported under the <code>Hpcc--*</code> ids.</li>
-<li><strong>Canned-cycle repeats do not take rotary words on SoftNc.</strong> The G73–G89 family expands XYZ / R /
-Q / F / P / K strokes. An A / B / C word on a repeat is not taken: a rotary-only repeat does not
-fire the cycle, a mixed block drops the rotary word, and under G68.2 the hole follows the table.
-<em>Not finished on this line, see Adoption status on the <a href="../index.html">release-note page</a>.</em> The
-HardNc fallback indexes those blocks the way a control does.</li>
+<li><strong>Canned-cycle repeats take rotary words on SoftNc.</strong> The G73–G89 family expands XYZ / R / Q /
+F / P / K strokes, and a repeat block carrying an A / B / C word indexes the table and drills
+again, the way a control does: a rotary-only repeat fires the cycle, a mixed block pre-positions
+XYZ and the table in one simultaneous move before the strokes, and under G68.2 without RTCP the
+feature frame stays pinned in machine space so the hole stays put while the workpiece turns. Under
+G91 only the words the block itself spoke are increments. The hold that 3.2.24 placed on these
+programs is lifted; <code>EnableSoftNcRunner = false</code> is no longer needed for them.</li>
+<li><strong>A rotary word under modal G91 is an increment on SoftNc.</strong> A bare A / B / C word follows the
+block's modal positioning onto the same anchor + delta path as Siemens <code>IC()</code> and klartext <code>IC+</code>:
+<code>G91 … C20.</code> then <code>C20.</code> leaves the table at 40. A per-word directive (<code>C=AC(90)</code>, <code>ACP</code> / <code>ACN</code>,
+<code>IC()</code>) still beats the modal state, and a spoken <code>G91 B270.</code> stays a +270° traverse instead of
+folding to −90°. A <code>G91 G28</code> rotary intermediate anchors on the previous modal angle, then
+returns home. HardNc keeps its absolute reading of a rotary word under G91 and warns. The
+programmed deltas are recorded in <a class="xref" href="../../api/Hi.NcParsers.Keywords.IncrementalWords.html">IncrementalWords</a>, beside the
+<a class="xref" href="../../api/Hi.NcParsers.Keywords.RotaryWords.html">RotaryWords</a> the block spoke.</li>
 </ul>
 <h2 id="cross-brand">Cross-brand</h2>
 <ul>
@@ -386,6 +386,13 @@ the authored one — the only face a project file persists, reached as
 physics, collision and execution display read; it is rebuilt from the authored face at project
 assignment and at session boundaries, so a value written onto it is discarded rather than saved.
 See <a href="../../technique/api-foundations/getting-started.html">Getting started</a>.</li>
+<li><code>BackgroundTemperature_C</code> / <code>_K</code> on both equipment faces rename to
+<a class="xref" href="../../api/Hi.Machining.MachiningEquipmentUtils.SetupEquipment.html#Hi_Machining_MachiningEquipmentUtils_SetupEquipment_AmbientTemperature_C">AmbientTemperature_C</a> /
+<a class="xref" href="../../api/Hi.Machining.MachiningEquipmentUtils.SetupEquipment.html#Hi_Machining_MachiningEquipmentUtils_SetupEquipment_AmbientTemperature_K">AmbientTemperature_K</a> (and the same pair
+on <a class="xref" href="../../api/Hi.Machining.MachiningEquipmentUtils.MachiningEquipment.html">MachiningEquipment</a>): the shop-floor temperature the
+workpiece, cutter, chips and spindle start at and the spindle cools toward. The old names remain
+as <code>[Obsolete]</code> delegates, so nothing stops compiling. A <code>.hincproj</code> now writes
+<code>&lt;AmbientTemperature_C&gt;</code>; the reader still accepts the legacy element and the 2019 wrapper.</li>
 </ul>
 <h2 id="8-defaults-and-gates-that-changed">8. Defaults and gates that changed</h2>
 <ul>
@@ -120,6 +120,22 @@ mirrors <code>RemoveBySweepingVolume</code> — the geometry-layer symmetric poi
 no deposition step in the machining runner yet, so this is a geometry-level API only.</li>
 <li><strong><a class="xref" href="../../api/Hi.Milling.FluteContours.FreeformBottomContour.html">FreeformBottomContour</a></strong> is the user-editable point-list bottom
 contour, mirroring the side contour with the key axis switched to radius.</li>
+<li><strong>A host with no usable OpenGL no longer aborts the process.</strong> The native display initialiser
+used to end the whole process when the GL loader failed — before any NC play, physics or training,
+none of which need GL — which is what made every service on a machine whose graphics driver and
+kernel module disagree unstartable, and kept it dead until the driver was fixed. It now reports
+and returns; <a class="xref" href="../../api/Hi.Disp.DispEngine.html#Hi_Disp_DispEngine_IsGlAvailable">IsGlAvailable</a> carries the verdict, <code>ContextProjDepth</code> reads
+0 in that state, and HiNc logs one line at startup saying so. The contract is that a host must
+not obtain a <a class="xref" href="../../api/Hi.Disp.DispEngine.html">DispEngine</a> while <code>IsGlAvailable</code> is <code>false</code>, rather than guard
+individual draw calls.</li>
+<li><strong>The cube tree folds a re-cut flat region again.</strong> A comparison in the simplification gate had
+been inverted since 2023: it demanded the marching-cubes ambiguous face pattern instead of
+forbidding it, so almost no cell reached the reproduction test and a floor roughened by one pass
+and re-cut flat by a later one stayed at its finest level for the life of the tree. 128 of the 256
+corner masks now pass, the 1e-9 reproduction test behind the gate is unchanged. On a 1000-move CL
+replay the node count drops 17% and the peak working set 8%. Because folds are now routine, the
+meshed-geometry writer and the fly-piece remover tolerate a child array folded away under them
+while a cut runs on another thread — a snapshot taken mid-cut stays torn, but no longer crashes.</li>
 </ul>

 </article>
@@ -88,7 +88,7 @@
          <h1 id="upgrading-from-31175-to-32">Upgrading from 3.1.175 to 3.2</h1>

 <p>These pages are the long form of the 3.1.175 → 3.2 line — the 3.2.7 entry and the
-<a href="../index.html">3.2.24</a> incremental drop. They exist because 3.2 is not
+<a href="../index.html">3.2.24 and 3.2.26</a> incremental drops. They exist because 3.2 is not
 an increment on the last release most callers hold — it is the accumulation of everything that
 landed after the 3.1.175 package set, delivered in one step.</p>
 <p>Read them in order the first time. The two sections that decide whether an upgrade is a recompile or
@@ -99,9 +99,9 @@ capability you can adopt when you need it.</p>
 <h5>Important</h5>
 <p>Read the <strong>Adoption status</strong> note on the <a href="../index.html">release-note page</a> before planning around this.
 The short version: 3.2 is published for review, verification is still in progress across most
-areas, <strong>NC optimization is not finished on this line</strong>, and <strong>SoftNc canned-cycle indexing
-with rotary A/B/C words is not finished</strong> — work that depends on either should stay on 3.1,
-serviced as <code>3.1.175.&lt;patch&gt;</code>, or run with SoftNc off for the rotary-cycle case.</p>
+areas and <strong>NC optimization is not finished on this line</strong> — work that depends on it should stay
+on 3.1, serviced as <code>3.1.175.&lt;patch&gt;</code>. The 3.2.24 hold on SoftNc canned cycles that index a
+rotary axis is lifted in 3.2.26.</p>
 </div>
 <h2 id="pages">Pages</h2>
 <p>Ordered the way to read them: what the package numbers mean, then the two sections that decide whether the upgrade is a recompile or an afternoon, then the new capability, adopted when needed.</p>
@@ -191,6 +191,10 @@ step touched the workpiece but no touched step carries a physics brief — namin
 gate it: a tool bound to the spindle, the spindle actually rotating, and the workpiece cutting
 parameter). That last state previously surfaced one process later, as a training run gathering zero
 samples.</li>
+<li><strong>Engagement is tessellation-invariant.</strong> The scan that turns a step's contact contours into
+the engaged angle ranges no longer depends on how many polygons the geometry kernel cut the
+contact surface into; forces on existing programs move accordingly — see
+<a href="results-that-change-on-upgrade.html">Results that change on upgrade</a>.</li>
 <li><strong>Performance</strong> is collected in its own section below.</li>
 </ul>

@@ -242,6 +242,31 @@ them on the bench fixtures); the largest moves are floors that used to register
 whole bottom disc dropped out — and side ranges that were fragmented into pieces separated by
 phantom 1° gaps. Coefficient sets trained before this change embed the old engagement of their
 training program; retrain on a replay if that fit matters.</li>
+<li><strong>SoftNc canned cycles index a rotary axis.</strong> A G81–G89 repeat block carrying an A / B / C word
+used to skip the hole (rotary-only repeat), drill it at the previous angle (mixed block) or let it
+follow the table under <code>G68.2</code>. It now indexes the table and drills again, pre-positioning XYZ and
+the table in one move, and under <code>G68.2</code> without RTCP the feature frame stays pinned in machine
+space. Under <code>G91</code> only the words the block itself spoke are increments — an X-pitch repeat used
+to add the previous cycle's Y a second time and drill upward. A bolt-circle program that ran
+short of holes, or drilled them in the wrong place, now matches the HardNc replay.</li>
+<li><strong>A rotary word under modal <code>G91</code> is an increment.</strong> <code>G91 … C20.</code> then <code>C20.</code> left the table at
+20; it is 40 now. A spoken <code>G91 B270.</code> stays a +270° traverse instead of folding to −90°, and a
+<code>G91 G28</code> rotary intermediate anchors on the previous modal angle instead of on zero (<code>G91 G28 C0</code> from C90 now holds the table and returns home instead of swinging to 0 first). Per-word
+directives (<code>C=AC(90)</code>, <code>ACP</code> / <code>ACN</code>, <code>IC()</code>) are unchanged. HardNc keeps its absolute reading
+and warns.</li>
+<li><strong>Rotary IK from a pole seed.</strong> Every solve attempt now runs coarse then polish, so a small tilt
+with a large azimuth that ran out of iterations a few milliradians short — and was refused by
+<code>G53.1</code>, leaving the table at A = C = 0 with the tilt active and the path rotated off the part —
+now solves to 1e-8 rad. On a Fanuc IJK grid from the pole seed 54 / 69 failures per A/C chain
+become 0; a CL program goes from 44 dropped endpoints of 536 to 0.</li>
+<li><strong>The rotary branch inside the travel wins.</strong> When the nearest IK solution lies outside an axis'
+configured travel, <code>G53.1</code>, <code>G68.2</code>, <code>CYCLE800</code>, <code>PLANE SPATIAL</code> and <code>LN</code> now take the admissible
+branch nearest the entry (sign, half-turn and supplement combinations are searched); a posture
+the cradle cannot take is no longer commanded silently. This needs the travel to be known, and
+it now is on every project: stroke limits declared only in the legacy machine-limits boxes are
+copied into the controller parameter table on load for the axes carrying no end there, where they
+used to be checked against nothing on SoftNc. An axis parked exactly on its limit is no longer
+reported (<a class="xref" href="../../api/Hi.NcParsers.Dependencys.StrokeLimitUtil.html#Hi_NcParsers_Dependencys_StrokeLimitUtil_Tolerance">Tolerance</a>).</li>
 <li>The SoftNc pipeline is the default engine.</li>
 <li>Executed <code>SyntaxPiece</code>s freeze to UTF-8 (below), so their <code>JsonObject</code> is a fresh read-only snapshot
 per call rather than a retained live graph.</li>
@@ -158,6 +158,13 @@ rewinds the NC-runner session state, so the next play restarts file and line ind
 the way <code>PowerReset</code> does, and it resets the CL device pose to identity. It no longer clobbers
 <code>MachiningResolution_mm</code>: the runtime resolution is seeded from the workpiece's initial resolution
 only when a project loads, so an explicit override survives a runtime reset and a workpiece swap.</li>
+<li><strong>Every NC-pipeline rewind raises one event.</strong> Suit rewiring, <code>PowerReset</code> and
+<a class="xref" href="../../api/Hi.MachiningProcs.SessionShell.html#Hi_MachiningProcs_SessionShell_ResetRuntime_">ResetRuntime</a> each dropped the runner session state on their
+own with no notification, so the file / sentence counters restarted at 0 on the next play while
+the session and its subscribers lived on — a consumer keyed by file index merged the next NC file
+into the previous file's numbers. One seam now performs the reset (session state and the NC
+diagnostics) and raises <a class="xref" href="../../api/Hi.MachiningProcs.LocalProjectService.html#Hi_MachiningProcs_LocalProjectService_OnNcPipelineReset">OnNcPipelineReset</a>, an
+app-lifetime event, before the next file's pieces arrive.</li>
 <li><strong>Mixed runner kinds in one session are refused.</strong> <code>NcRunnerSessionState</code> remembers the runner that
 initialised it and <a class="xref" href="../../api/Hi.NcParsers.SoftNcRunner.html#Hi_NcParsers_SoftNcRunner_RunNcLines_">RunNcLines</a> refuses a different one with
 <code>RunNcLines--RunnerMismatch</code> — reachable now that <code>NcKind.Auto</code> makes mixed-kind missions a