Suspense: Active → Suspended → Resolved with fallback (M1-T15, closes #23)

[ncdevshiv]

What

Closes#23. The Suspense lifecycle with a real (not faked) pending source:

constres=useResource(0);// (Pending, resolve)<Suspensefallback={<p>loading</p>}><p>{res[0]}</p></Suspense>

• Value::Pending + useResource(key): a state-driven source — reads are Pending until the resolver Setter fires (dirty → flush → re-render). No timers or fakes.
• ReactNode::Suspense: the Text arm turns a Pending read into a RenderedNode::Suspended marker; the Suspense arm scans recursively (a Pending text inside a host child must suspend — direct-children-only missed it) and swaps the subtree for the fallback.
• Resolve is minimal: one SetProp + one SetText, zero removes/creates — the swap is a branch in the same logical position, not a tree rewrite (test-verified).
• Resolved state sticks across unrelated re-renders; two boundaries resolve independently (per-instance).
• Bug the smoke caught: use_pending always returned Pending, so resolve was a no-op; it now returns the stored slot value.

Verification

4 tests in tests/suspense.rs: fallback→resolve with minimal patches, siblings render while suspended, per-instance independence, unresolved stays suspended across unrelated re-renders. Suite 149 green; clippy -D warnings clean; fmt clean; audit green. Records: M1 15/18, overall 51/106.

[ncdevshiv]

What

Closes#23. The Suspense lifecycle with a real (not faked) pending source:

constres=useResource(0);// (Pending, resolve)<Suspensefallback={<p>loading</p>}><p>{res[0]}</p></Suspense>

• Value::Pending + useResource(key): a state-driven source — reads are Pending until the resolver Setter fires (dirty → flush → re-render). No timers or fakes.
• ReactNode::Suspense: the Text arm turns a Pending read into a RenderedNode::Suspended marker; the Suspense arm scans recursively (a Pending text inside a host child must suspend — direct-children-only missed it) and swaps the subtree for the fallback.
• Resolve is minimal: one SetProp + one SetText, zero removes/creates — the swap is a branch in the same logical position, not a tree rewrite (test-verified).
• Resolved state sticks across unrelated re-renders; two boundaries resolve independently (per-instance).
• Bug the smoke caught: use_pending always returned Pending, so resolve was a no-op; it now returns the stored slot value.

Verification

4 tests in tests/suspense.rs: fallback→resolve with minimal patches, siblings render while suspended, per-instance independence, unresolved stays suspended across unrelated re-renders. Suite 149 green; clippy -D warnings clean; fmt clean; audit green. Records: M1 15/18, overall 51/106.

[ncdevshiv]

What

Closes#23. The Suspense lifecycle with a real (not faked) pending source:

constres=useResource(0);// (Pending, resolve)<Suspensefallback={<p>loading</p>}><p>{res[0]}</p></Suspense>

• Value::Pending + useResource(key): a state-driven source — reads are Pending until the resolver Setter fires (dirty → flush → re-render). No timers or fakes.
• ReactNode::Suspense: the Text arm turns a Pending read into a RenderedNode::Suspended marker; the Suspense arm scans recursively (a Pending text inside a host child must suspend — direct-children-only missed it) and swaps the subtree for the fallback.
• Resolve is minimal: one SetProp + one SetText, zero removes/creates — the swap is a branch in the same logical position, not a tree rewrite (test-verified).
• Resolved state sticks across unrelated re-renders; two boundaries resolve independently (per-instance).
• Bug the smoke caught: use_pending always returned Pending, so resolve was a no-op; it now returns the stored slot value.

Verification

4 tests in tests/suspense.rs: fallback→resolve with minimal patches, siblings render while suspended, per-instance independence, unresolved stays suspended across unrelated re-renders. Suite 149 green; clippy -D warnings clean; fmt clean; audit green. Records: M1 15/18, overall 51/106.

[ncdevshiv]

What

Closes#23. The Suspense lifecycle with a real (not faked) pending source:

constres=useResource(0);// (Pending, resolve)<Suspensefallback={<p>loading</p>}><p>{res[0]}</p></Suspense>

• Value::Pending + useResource(key): a state-driven source — reads are Pending until the resolver Setter fires (dirty → flush → re-render). No timers or fakes.
• ReactNode::Suspense: the Text arm turns a Pending read into a RenderedNode::Suspended marker; the Suspense arm scans recursively (a Pending text inside a host child must suspend — direct-children-only missed it) and swaps the subtree for the fallback.
• Resolve is minimal: one SetProp + one SetText, zero removes/creates — the swap is a branch in the same logical position, not a tree rewrite (test-verified).
• Resolved state sticks across unrelated re-renders; two boundaries resolve independently (per-instance).
• Bug the smoke caught: use_pending always returned Pending, so resolve was a no-op; it now returns the stored slot value.

Verification

4 tests in tests/suspense.rs: fallback→resolve with minimal patches, siblings render while suspended, per-instance independence, unresolved stays suspended across unrelated re-renders. Suite 149 green; clippy -D warnings clean; fmt clean; audit green. Records: M1 15/18, overall 51/106.

[ncdevshiv]

What

Closes#23. The Suspense lifecycle with a real (not faked) pending source:

constres=useResource(0);// (Pending, resolve)<Suspensefallback={<p>loading</p>}><p>{res[0]}</p></Suspense>

• Value::Pending + useResource(key): a state-driven source — reads are Pending until the resolver Setter fires (dirty → flush → re-render). No timers or fakes.
• ReactNode::Suspense: the Text arm turns a Pending read into a RenderedNode::Suspended marker; the Suspense arm scans recursively (a Pending text inside a host child must suspend — direct-children-only missed it) and swaps the subtree for the fallback.
• Resolve is minimal: one SetProp + one SetText, zero removes/creates — the swap is a branch in the same logical position, not a tree rewrite (test-verified).
• Resolved state sticks across unrelated re-renders; two boundaries resolve independently (per-instance).
• Bug the smoke caught: use_pending always returned Pending, so resolve was a no-op; it now returns the stored slot value.

Verification

4 tests in tests/suspense.rs: fallback→resolve with minimal patches, siblings render while suspended, per-instance independence, unresolved stays suspended across unrelated re-renders. Suite 149 green; clippy -D warnings clean; fmt clean; audit green. Records: M1 15/18, overall 51/106.

[ncdevshiv]

What

Closes#23. The Suspense lifecycle with a real (not faked) pending source:

constres=useResource(0);// (Pending, resolve)<Suspensefallback={<p>loading</p>}><p>{res[0]}</p></Suspense>

• Value::Pending + useResource(key): a state-driven source — reads are Pending until the resolver Setter fires (dirty → flush → re-render). No timers or fakes.
• ReactNode::Suspense: the Text arm turns a Pending read into a RenderedNode::Suspended marker; the Suspense arm scans recursively (a Pending text inside a host child must suspend — direct-children-only missed it) and swaps the subtree for the fallback.
• Resolve is minimal: one SetProp + one SetText, zero removes/creates — the swap is a branch in the same logical position, not a tree rewrite (test-verified).
• Resolved state sticks across unrelated re-renders; two boundaries resolve independently (per-instance).
• Bug the smoke caught: use_pending always returned Pending, so resolve was a no-op; it now returns the stored slot value.

Verification

4 tests in tests/suspense.rs: fallback→resolve with minimal patches, siblings render while suspended, per-instance independence, unresolved stays suspended across unrelated re-renders. Suite 149 green; clippy -D warnings clean; fmt clean; audit green. Records: M1 15/18, overall 51/106.

[ncdevshiv]

What

Closes#23. The Suspense lifecycle with a real (not faked) pending source:

constres=useResource(0);// (Pending, resolve)<Suspensefallback={<p>loading</p>}><p>{res[0]}</p></Suspense>

• Value::Pending + useResource(key): a state-driven source — reads are Pending until the resolver Setter fires (dirty → flush → re-render). No timers or fakes.
• ReactNode::Suspense: the Text arm turns a Pending read into a RenderedNode::Suspended marker; the Suspense arm scans recursively (a Pending text inside a host child must suspend — direct-children-only missed it) and swaps the subtree for the fallback.
• Resolve is minimal: one SetProp + one SetText, zero removes/creates — the swap is a branch in the same logical position, not a tree rewrite (test-verified).
• Resolved state sticks across unrelated re-renders; two boundaries resolve independently (per-instance).
• Bug the smoke caught: use_pending always returned Pending, so resolve was a no-op; it now returns the stored slot value.

Verification

4 tests in tests/suspense.rs: fallback→resolve with minimal patches, siblings render while suspended, per-instance independence, unresolved stays suspended across unrelated re-renders. Suite 149 green; clippy -D warnings clean; fmt clean; audit green. Records: M1 15/18, overall 51/106.

[ncdevshiv]

What

Closes#23. The Suspense lifecycle with a real (not faked) pending source:

constres=useResource(0);// (Pending, resolve)<Suspensefallback={<p>loading</p>}><p>{res[0]}</p></Suspense>

• Value::Pending + useResource(key): a state-driven source — reads are Pending until the resolver Setter fires (dirty → flush → re-render). No timers or fakes.
• ReactNode::Suspense: the Text arm turns a Pending read into a RenderedNode::Suspended marker; the Suspense arm scans recursively (a Pending text inside a host child must suspend — direct-children-only missed it) and swaps the subtree for the fallback.
• Resolve is minimal: one SetProp + one SetText, zero removes/creates — the swap is a branch in the same logical position, not a tree rewrite (test-verified).
• Resolved state sticks across unrelated re-renders; two boundaries resolve independently (per-instance).
• Bug the smoke caught: use_pending always returned Pending, so resolve was a no-op; it now returns the stored slot value.

Verification

4 tests in tests/suspense.rs: fallback→resolve with minimal patches, siblings render while suspended, per-instance independence, unresolved stays suspended across unrelated re-renders. Suite 149 green; clippy -D warnings clean; fmt clean; audit green. Records: M1 15/18, overall 51/106.