fix: ODT sentinel passes anchor at the block tag, effects ride past annotations

A text box's anchor paragraph nests its children, so the STY/PBX passes'
scan to the closing tag started at the anchor and stole the first box
child's spec (its indent, direction, shading and borders). Both passes
now match the sentinel right after the opening tag, where it is emitted.
The TEF fallback also accepts bookmark markers and comment annotations
inside a run (tempered to the annotation's first close tag, -end kept
distinct), instead of cutting the effect short at them.

src/lib/export/odt.ts

@@ -1931,19 +1931,18 @@ function applyParagraphStyles(odtBytes: Uint8Array): Uint8Array {
     return name;
   };
 
-  const styRe = new RegExp(`${STY}([^${STY}]*)${STY}`);
+  // Anchored right after the opening tag (where the sentinel is emitted): a scan to the
+  // closing tag would start at a text box's anchor paragraph and steal the first box
+  // child's sentinel — nested <text:p> (box children, note bodies) make spans unsafe.
   content = content.replace(
-    new RegExp(`<text:(p|h)\\b((?:[^>]*[^/>])?)>([\\s\\S]*?)</text:\\1>`, 'g'),
-    (m, tag: string, attrs: string, inner: string) => {
-      const sm = styRe.exec(inner);
-      if (!sm) return m;
-      const cleaned = inner.replace(styRe, '');
+    new RegExp(`<text:(p|h)\\b((?:[^>]*[^/>])?)>${STY}([^${STY}]*)${STY}`, 'g'),
+    (_m, tag: string, attrs: string, styleName: string) => {
       const srcM = /text:style-name="([^"]*)"/.exec(attrs);
-      const name = styleFor(srcM ? srcM[1] : '', sm[1]);
+      const name = styleFor(srcM ? srcM[1] : '', styleName);
       const newAttrs = srcM
         ? attrs.replace(/text:style-name="[^"]*"/, `text:style-name="${name}"`)
         : ` text:style-name="${name}"${attrs}`;
-      return `<text:${tag}${newAttrs}>${cleaned}</text:${tag}>`;
+      return `<text:${tag}${newAttrs}>`;
     },
   );
   content = content.replace(new RegExp(`${STY}[^${STY}]*${STY}`, 'g'), '');
@@ -1989,19 +1988,18 @@ function applyParagraphBoxes(odtBytes: Uint8Array): Uint8Array {
     return name;
   };
 
-  const pbxRe = new RegExp(`${PBX}([^${PBX}]*)${PBX}`);
+  // Anchored right after the opening tag, over an optional STY run (its pass runs later;
+  // emission order is FSZ STY PBX). A scan to the closing tag would start at a text box's
+  // anchor paragraph and steal the first box child's sentinel.
   content = content.replace(
-    new RegExp(`<text:(p|h)\\b((?:[^>]*[^/>])?)>([\\s\\S]*?)</text:\\1>`, 'g'),
-    (m, tag: string, attrs: string, inner: string) => {
-      const sm = pbxRe.exec(inner);
-      if (!sm) return m;
-      const cleaned = inner.replace(pbxRe, '');
+    new RegExp(`<text:(p|h)\\b((?:[^>]*[^/>])?)>((?:${STY}[^${STY}]*${STY})?)${PBX}([^${PBX}]*)${PBX}`, 'g'),
+    (_m, tag: string, attrs: string, sty: string, spec: string) => {
       const srcM = /text:style-name="([^"]*)"/.exec(attrs);
-      const name = boxStyleFor(srcM ? srcM[1] : '', sm[1]);
+      const name = boxStyleFor(srcM ? srcM[1] : '', spec);
       const newAttrs = srcM
         ? attrs.replace(/text:style-name="[^"]*"/, `text:style-name="${name}"`)
         : ` text:style-name="${name}"${attrs}`;
-      return `<text:${tag}${newAttrs}>${cleaned}</text:${tag}>`;
+      return `<text:${tag}${newAttrs}>${sty}`;
     },
   );
   // Drop any sentinels not consumed above (defensive — never legitimate text).
@@ -3191,10 +3189,17 @@ function resolveTextEffects(xml: string, mint: (styleXml: string) => void, prefi
     },
   );
   // A run whose only formatting is such an effect never got a span: give it one, up to
-  // the next real element — a tab/line-break/space inside the run rides along, or the
-  // text behind it would fall out of the span and lose the effect.
+  // the next real element — a tab/line-break/space, a bookmark marker or a comment's
+  // annotation inside the run rides along, or the text behind it would fall out of the
+  // span and lose the effect.
+  // The annotation block is tempered to its first close tag, and (?=[\s>]) keeps it off
+  // <office:annotation-end (\b sits before the hyphen) — either slip would stretch one
+  // annotation across the next one's body and misplace the minted span's close.
+  const inline = `<text:(?:tab|line-break|s|bookmark-start|bookmark-end)\\b[^>]*/>`
+    + `|<office:annotation(?=[\\s>])(?:(?!</office:annotation>)[\\s\\S])*</office:annotation>`
+    + `|<office:annotation-end\\b[^>]*/>`;
   out = out.replace(
-    new RegExp(`${TEF}([^${TEF}]+)${TEF}((?:[^<${TEF}]|<text:(?:tab|line-break|s)\\b[^>]*/>)*)(?:${TEF}${TEF})?`, 'g'),
+    new RegExp(`${TEF}([^${TEF}]+)${TEF}((?:[^<${TEF}]|${inline})*)(?:${TEF}${TEF})?`, 'g'),
     (_m, extra: string, text: string) => `<text:span text:style-name="${styleFor('', extra)}">${text}</text:span>`,
   );
   // Drop any sentinels not consumed above (defensive — never legitimate text).