feat: updated text handling bugs

CMakePresets.json

@@ -119,7 +119,8 @@
       "cacheVariables": {
         "CMAKE_BUILD_TYPE": "Release",
         "PDFENGINE_BUILD_TESTS": "OFF",
-        "PDFENGINE_WITH_PDFIUM": "ON"
+        "PDFENGINE_WITH_PDFIUM": "ON",
+        "PDFENGINE_WITH_QPDF": "ON"
       }
     },
 

bindings/python/pdfengine_py.cpp

@@ -492,5 +492,9 @@ PYBIND11_MODULE(pdfengine, m) {
         .def("save_full", [](const pdfengine::PdfDocument& self) {
             std::vector<uint8_t> res = get_or_throw(self.saveFull());
             return py::bytes(reinterpret_cast<const char*>(res.data()), res.size());
+        })
+        .def("save_full_for_export", [](const pdfengine::PdfDocument& self) {
+            std::vector<uint8_t> res = get_or_throw(self.saveFullForExport());
+            return py::bytes(reinterpret_cast<const char*>(res.data()), res.size());
         });
 }

engine/include/pdfengine/pdf_document.hpp

@@ -265,8 +265,11 @@ public:
     [[nodiscard]] virtual std::expected<std::vector<uint8_t>, EngineError> 
     saveIncremental() const = 0;
 
-    [[nodiscard]] virtual std::expected<std::vector<uint8_t>, EngineError> 
+    [[nodiscard]] virtual std::expected<std::vector<uint8_t>, EngineError>
     saveFull() const = 0;
+
+    [[nodiscard]] virtual std::expected<std::vector<uint8_t>, EngineError>
+    saveFullForExport() const { return saveFull(); }
 };
 
 }

engine/src/fonts/shaping/hb_shaper.cpp

@@ -3,6 +3,9 @@
 #include <hb.h>
 #include <hb-ft.h>
 #include <mutex>
+#include <spdlog/spdlog.h>
+#include <ft2build.h>
+#include FT_FREETYPE_H
 
 namespace pdfengine::fonts {
 
@@ -25,8 +28,13 @@ std::vector<ShapedGlyph> HbShaper::shapeRun(
     std::lock_guard<std::mutex> lock(font.getMutex());
 
     if (FT_Set_Pixel_Sizes(ftFace, 0, fontSize)) {
+        spdlog::warn("HbShaper: FT_Set_Pixel_Sizes({}) failed (numFixedSizes={} scalable={})",
+                     fontSize, ftFace->num_fixed_sizes, (FT_IS_SCALABLE(ftFace) ? 1 : 0));
         return result;
     }
+    if (FT_Select_Charmap(ftFace, FT_ENCODING_UNICODE) != 0 && ftFace->num_charmaps > 0) {
+        FT_Set_Charmap(ftFace, ftFace->charmaps[0]);
+    }
 
     hb_font_t* hbFont = hb_ft_font_create_referenced(ftFace);
     if (!hbFont) {

engine/src/parser/pdfium_document.cpp

@@ -1,5 +1,9 @@
 #include "parser/pdfium_internal.hpp"
 
+#if defined(PDFENGINE_WITH_PDFIUM) && defined(PDFENGINE_WITH_QPDF)
+#include "qpdf/qpdf_writer.hpp"
+#endif
+
 namespace pdfengine {
 
 std::expected<std::shared_ptr<PdfDocument>, EngineError> 
@@ -240,6 +244,20 @@ std::expected<std::vector<uint8_t>, EngineError> PdfiumDocument::saveFull() cons
 #endif
 }
 
+std::expected<std::vector<uint8_t>, EngineError> PdfiumDocument::saveFullForExport() const {
+    auto bytes = saveFull();
+#if defined(PDFENGINE_WITH_PDFIUM) && defined(PDFENGINE_WITH_QPDF)
+    if (bytes) {
+        qpdf_layer::QpdfWriter writer;
+        auto withAppearances = writer.setNeedAppearances(*bytes);
+        if (withAppearances) return *withAppearances;
+        spdlog::warn("saveFullForExport: NeedAppearances pass failed ({}); returning plain save",
+                     withAppearances.error());
+    }
+#endif
+    return bytes;
+}
+
 void PdfiumDocument::invalidateCaches() {
     {
         std::lock_guard<std::mutex> lock(fontsMutex_);

engine/src/parser/pdfium_document.hpp

@@ -103,6 +103,7 @@ public:
     std::expected<void, EngineError> applyEdits(const std::string& editsJson) override;
     std::expected<std::vector<uint8_t>, EngineError> saveIncremental() const override;
     std::expected<std::vector<uint8_t>, EngineError> saveFull() const override;
+    std::expected<std::vector<uint8_t>, EngineError> saveFullForExport() const override;
 
     std::string lastReflowLayout() const { return lastReflowLayout_; }
     bool lastReflowOverflowed() const { return lastReflowOverflowed_; }
@@ -158,6 +159,8 @@ private:
     const fonts::pdf_fonts::ReconstructedFont* getReconstructedEmbeddedFont(const std::string& internalFontId);
 #endif
 
+    void rebalanceEditedPages(const std::vector<int>& editedPages);
+
     std::expected<void, EngineError> applyOp_replaceText(const nlohmann::json& op, int pageIndex);
     std::expected<void, EngineError> applyOp_reflow(const nlohmann::json& op, int pageIndex);
     std::expected<void, EngineError> applyOp_textOverlay(const nlohmann::json& op, int pageIndex);

engine/src/parser/pdfium_edit.cpp

@@ -1,5 +1,11 @@
 #include "parser/pdfium_internal.hpp"
 
+#include <set>
+#include <algorithm>
+#if defined(PDFENGINE_WITH_PDFIUM) && defined(PDFENGINE_WITH_QPDF)
+#include "qpdf/qpdf_writer.hpp"
+#endif
+
 namespace pdfengine::parser {
 
 std::expected<void, EngineError> PdfiumDocument::applyEdits(const std::string& editsJson) {
@@ -10,6 +16,12 @@ std::expected<void, EngineError> PdfiumDocument::applyEdits(const std::string& e
 
     lastReflowLayout_.clear();
 
+    std::vector<int> editedPages;
+    auto markEdited = [&editedPages](int p) {
+        if (std::find(editedPages.begin(), editedPages.end(), p) == editedPages.end())
+            editedPages.push_back(p);
+    };
+
     try {
         auto root = nlohmann::json::parse(editsJson);
         if (!root.contains("version") || root["version"] != "1.0") {
@@ -29,6 +41,12 @@ std::expected<void, EngineError> PdfiumDocument::applyEdits(const std::string& e
                 return std::unexpected(EngineError::PageOutOfBounds);
             }
 
+            static const std::set<std::string> kContentOps = {
+                "replace_text", "reflow_paragraph", "text_overlay", "add_text",
+                "underline", "strikeout", "squiggly", "redaction",
+                "image_overlay", "highlight", "free_text", "comment", "freehand"};
+            if (kContentOps.count(type)) markEdited(pageIndex);
+
             std::expected<void, EngineError> r{};
             if (type == "replace_text") {
                 r = applyOp_replaceText(op, pageIndex);
@@ -78,6 +96,7 @@ std::expected<void, EngineError> PdfiumDocument::applyEdits(const std::string& e
         return std::unexpected(EngineError::Unknown);
     }
 
+    rebalanceEditedPages(editedPages);
     invalidateCaches();
     return {};
 #else
@@ -86,4 +105,34 @@ std::expected<void, EngineError> PdfiumDocument::applyEdits(const std::string& e
 #endif
 }
 
+void PdfiumDocument::rebalanceEditedPages(const std::vector<int>& editedPages) {
+#if defined(PDFENGINE_WITH_PDFIUM) && defined(PDFENGINE_WITH_QPDF)
+    if (!doc_ || editedPages.empty()) return;
+
+    auto bytesRes = saveFull();
+    if (!bytesRes || bytesRes->empty()) return;
+
+    qpdf_layer::QpdfWriter writer;
+    auto reb = writer.rebalanceContentStreams(*bytesRes, editedPages);
+    if (!reb) {
+        spdlog::warn("rebalanceEditedPages: {}", reb.error());
+        return;
+    }
+    if (!reb->has_value()) return; 
+
+    std::vector<uint8_t> newBytes = std::move(**reb);
+    FPDF_DOCUMENT nd = FPDF_LoadMemDocument(newBytes.data(), static_cast<int>(newBytes.size()), nullptr);
+    if (!nd) {
+        spdlog::warn("rebalanceEditedPages: reload of repaired doc failed; keeping original");
+        return;
+    }
+    FPDF_CloseDocument(doc_);
+    doc_ = nd;
+    memoryBuffer_ = std::move(newBytes);
+    spdlog::info("rebalanceEditedPages: repaired q/Q underflow on edited page(s) and reloaded doc");
+#else
+    (void)editedPages;
+#endif
+}
+
 }

engine/src/parser/pdfium_edit_reflow.cpp

@@ -67,6 +67,7 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_reflow(const nlohmann::
                     for (const auto& v : data["lineX"]) lineX.push_back(v.get<double>());
                 double columnWidth = columnRight - columnLeft;
                 double pushColumnLeft = data.value("pushColumnLeft", columnLeft);
+                double hangingIndent = data.value("hangingIndent", 0.0);
                 std::string paraId;
                 if (data.contains("paraId") && data["paraId"].is_string())
                     paraId = data["paraId"].get<std::string>();
@@ -278,6 +279,7 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_reflow(const nlohmann::
                 };
 
                 auto isSpace = [](char ch) { return ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r'; };
+                constexpr size_t kHardBreak = static_cast<size_t>(-1);
                 struct Seg { int runIdx; std::string text; double width; size_t off; };
                 struct Word { std::vector<Seg> segs; double width; double spaceAfter; };
                 std::vector<Word> words;
@@ -291,42 +293,68 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_reflow(const nlohmann::
                     std::vector<size_t> out;
                     size_t i = 0;
                     while (i < ft.size()) {
-                        if (isSpace(ft[i])) { i++; continue; }
+                        if (isSpace(ft[i])) { if (ft[i] == '\n') out.push_back(kHardBreak); i++; continue; }
                         Word w; w.width = 0.0; w.spaceAfter = 0.0;
                         while (i < ft.size() && !isSpace(ft[i])) {
                             int st = sof[i]; size_t segOff = cof[i]; std::string frag; double fw = 0.0;
                             while (i < ft.size() && !isSpace(ft[i]) && sof[i] == st) { frag.push_back(ft[i]); fw += charAdvAt(st, cof[i]); i++; }
                             w.segs.push_back({st, frag, fw, segOff}); w.width += fw;
                         }
-                        while (i < ft.size() && isSpace(ft[i])) { w.spaceAfter += charAdvAt(sof[i], cof[i]); i++; }
+                        size_t newlines = 0;
+                        while (i < ft.size() && isSpace(ft[i])) {
+                            if (ft[i] == '\n') newlines++;
+                            else if (newlines == 0) w.spaceAfter += charAdvAt(sof[i], cof[i]);
+                            i++;
+                        }
                         out.push_back(words.size());
                         words.push_back(std::move(w));
+                        for (size_t n = 0; n < newlines; ++n) out.push_back(kHardBreak);
                     }
                     return out;
                 };
 
                 std::vector<std::vector<size_t>> lines;
+                std::vector<char> lineCont;  
+                std::vector<char> lineSegEnd; 
                 if (hasProvidedLines) {
                     for (const auto& lineRunIdxs : providedLines) {
                         auto wi = tokenize(lineRunIdxs);
-                        if (!wi.empty()) lines.push_back(std::move(wi));
+                        wi.erase(std::remove(wi.begin(), wi.end(), kHardBreak), wi.end());
+                        if (!wi.empty()) { lines.push_back(std::move(wi)); lineCont.push_back(0); lineSegEnd.push_back(0); }
                     }
+                    for (size_t li = 0; li < lineSegEnd.size(); ++li)
+                        lineSegEnd[li] = (li + 1 == lineSegEnd.size()) ? 1 : 0;
                 } else {
                     std::vector<int> allRuns(runs.size());
                     for (size_t ri = 0; ri < runs.size(); ++ri) allRuns[ri] = static_cast<int>(ri);
                     auto allWords = tokenize(allRuns);
                     std::vector<size_t> cur; double curW = 0.0;
+                    size_t prevWord = kHardBreak;
+                    bool firstLineOfSeg = true;
+                    auto pushLine = [&](char segEnd) {
+                        lines.push_back(cur);
+                        lineCont.push_back(firstLineOfSeg ? 0 : 1);
+                        lineSegEnd.push_back(segEnd);
+                    };
                     for (size_t k = 0; k < allWords.size(); ++k) {
                         size_t wi = allWords[k];
-                        double gap = cur.empty() ? 0.0 : words[allWords[k - 1]].spaceAfter;
-                        if (!cur.empty() && curW + gap + words[wi].width > columnWidth) {
-                            lines.push_back(cur); cur.clear();
+                        if (wi == kHardBreak) {       // forced break: end the current line (may be blank)
+                            pushLine(1);
+                            cur.clear(); curW = 0.0; prevWord = kHardBreak; firstLineOfSeg = true;
+                            continue;
+                        }
+                        double effW = columnWidth - (firstLineOfSeg ? 0.0 : hangingIndent);
+                        double gap = (cur.empty() || prevWord == kHardBreak) ? 0.0 : words[prevWord].spaceAfter;
+                        if (!cur.empty() && curW + gap + words[wi].width > effW) {
+                            pushLine(0);              
+                            cur.clear(); firstLineOfSeg = false;
                             cur.push_back(wi); curW = words[wi].width;
                         } else {
                             cur.push_back(wi); curW += gap + words[wi].width;
                         }
+                        prevWord = wi;
                     }
-                    if (!cur.empty()) lines.push_back(cur);
+                    if (!cur.empty()) pushLine(1);
                 }
                 if (words.empty() || lines.empty()) { FPDF_ClosePage(page); return {}; }
                 int newLineCount = static_cast<int>(lines.size());
@@ -369,17 +397,20 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_reflow(const nlohmann::
                         naturalW += words[lw[k]].width;
                         if (k > 0) naturalW += words[lw[k - 1]].spaceAfter;
                     }
-                    bool justifyThis = (align == "justify") && (li + 1 < lines.size()) && lw.size() > 1;
+                    double indent = (li < lineCont.size() && lineCont[li]) ? hangingIndent : 0.0;
+                    double effCol = columnWidth - indent;
+                    bool segEnd = (li < lineSegEnd.size()) ? (lineSegEnd[li] != 0) : (li + 1 == lines.size());
+                    bool justifyThis = (align == "justify") && !segEnd && lw.size() > 1;
                     double extraPerGap = 0.0;
-                    if (justifyThis) { double slack = columnWidth - naturalW; if (slack > 0) extraPerGap = slack / static_cast<double>(lw.size() - 1); }
+                    if (justifyThis) { double slack = effCol - naturalW; if (slack > 0) extraPerGap = slack / static_cast<double>(lw.size() - 1); }
 
                     std::string lineText;
                     std::vector<double> adv;
                     double lineFontSize = 0.0;
-                    double x = (li < lineX.size()) ? lineX[li] : columnLeft;
+                    double x = (li < lineX.size()) ? lineX[li] : (columnLeft + indent);
                     if (li >= lineX.size()) {
-                        if (align == "right") x = columnLeft + (columnWidth - naturalW);
-                        else if (align == "center") x = columnLeft + (columnWidth - naturalW) / 2.0;
+                        if (align == "right") x = columnLeft + indent + (effCol - naturalW);
+                        else if (align == "center") x = columnLeft + indent + (effCol - naturalW) / 2.0;
                     }
                     const double lineStartX = x;
                     for (size_t k = 0; k < lw.size(); ++k) {

engine/src/parser/pdfium_edit_replace.cpp

@@ -176,7 +176,22 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
 
                 bool fontSupportsAll = true;
                 double totalWidth = 0.0;
-                
+
+                std::shared_ptr<fonts::FontFace> measureFace;
+                {
+                    std::vector<uint8_t> mb;
+                    if (auto perObj = getFontDataFromObjects(pageIndex, objectIndices, internalFontId);
+                        perObj.has_value() && !perObj->empty()) mb = std::move(*perObj);
+                    else if (!internalFontId.empty()) {
+                        if (auto fd = getFontData(internalFontId); fd.has_value() && !fd->empty())
+                            mb = std::move(fd.value());
+                    }
+                    if (!mb.empty()) {
+                        auto mf = std::make_shared<fonts::FontFace>();
+                        if (mf->loadFromMemory(mb)) measureFace = mf;
+                    }
+                }
+
                 auto utf16 = utf8_to_utf16le(newText);
                 std::vector<uint32_t> unicodeCodepoints;
                 for (size_t i = 0; i < utf16.size(); ) {
@@ -214,23 +229,25 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                 }
 
                 bool shapedSuccessful = false;
-                if (resolvedFont) {
+                constexpr unsigned int kRefMeasure = 1000;
+                fonts::FontFace* mface = measureFace ? measureFace.get()
+                                       : (resolvedFont ? &resolvedFont->getFontFace() : nullptr);
+                if (mface) {
                     try {
                         fonts::HbShaper shaper;
-                        unsigned int uFontSize = static_cast<unsigned int>(fontSize > 0.0 ? fontSize : 12.0);
-                        auto shapedGlyphs = shaper.shapeRun(newText, resolvedFont->getFontFace(), uFontSize);
+                        auto shapedGlyphs = shaper.shapeRun(newText, *mface, kRefMeasure);
                         if (!shapedGlyphs.empty()) {
-                            totalWidth = 0.0;
-                            for (const auto& sg : shapedGlyphs) {
-                                totalWidth += sg.advanceX;
-                            }
+                            double sum = 0.0;
+                            for (const auto& sg : shapedGlyphs) sum += sg.advanceX;
+                            totalWidth = sum * (fontSize > 0.0 ? fontSize : 12.0) / static_cast<double>(kRefMeasure);
                             shapedSuccessful = true;
-                            spdlog::info("Font Engine: HarfBuzz shaped '{}' glyphs, total advance width = {}", shapedGlyphs.size(), totalWidth);
+                            spdlog::info("Font Engine: measured {} chars -> width {:.2f} ({} face)",
+                                         newText.size(), totalWidth, measureFace ? "embedded" : "resolved");
                         }
                     } catch (const std::exception& e) {
-                        spdlog::warn("Font Engine: HarfBuzz shaping failed: {}", e.what());
+                        spdlog::warn("Font Engine: shaping failed: {}", e.what());
                     } catch (...) {
-                        spdlog::warn("Font Engine: HarfBuzz shaping failed with unknown exception");
+                        spdlog::warn("Font Engine: shaping failed (unknown)");
                     }
                 }
 
@@ -269,7 +286,9 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
 
                 bool axisAligned = (std::abs(b) < 1e-6 && std::abs(c) < 1e-6 && a > 0.0 && d > 0.0);
                 double colRight = origRight;
-                bool sawSibling = false; 
+                bool sawSibling = false;
+                bool hasFollowingText = false;
+                double rowTolFollow = (std::max)(4.0, (origTop - origBottom) * 0.6);
                 if (axisAligned && hasOrigBounds) {
                     int nObjForCol = FPDFPage_CountObjects(page);
                     for (int k = 0; k < nObjForCol; ++k) {
@@ -279,44 +298,16 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                         float l = 0, bo = 0, rr = 0, tt = 0;
                         if (!FPDFPageObj_GetBounds(o, &l, &bo, &rr, &tt)) continue;
                         if (std::abs(l - origLeft) <= 3.0) { sawSibling = true; if (rr > colRight) colRight = rr; }
+                        if (std::abs((bo + tt) / 2.0 - origCenterY) <= rowTolFollow && l > origRight + 1.0)
+                            hasFollowingText = true;
                     }
                 }
                 double justifyTol = (std::max)(4.0, (colRight - origLeft) * 0.02);
                 bool wasJustified = !disableJustify && axisAligned && resolvedFont && hasOrigBounds && sawSibling &&
+                                    !hasFollowingText &&
                                     (colRight - origLeft) > 20.0 && (origRight >= colRight - justifyTol);
 
-                double deltaX = 0.0;
-                if (hasOrigBounds) {
-                    deltaX = totalWidth - origWidth;
-                    spdlog::info("Reflow Engine: origWidth = {}, newWidth = {}, deltaX = {}", origWidth, totalWidth, deltaX);
-                }
-
-                if (!wasJustified && !disableJustify && hasOrigBounds && std::abs(deltaX) > 0.001) {
-                    int pageObjCount = FPDFPage_CountObjects(page);
-                    double tolerance = (std::max)(5.0, fontSize * 0.5);
-                    int reflowedCount = 0;
-                    
-                    for (int k = 0; k < pageObjCount; ++k) {
-                        if (std::find(objectIndices.begin(), objectIndices.end(), k) != objectIndices.end()) {
-                            continue;
-                        }
-                        
-                        FPDF_PAGEOBJECT otherObj = FPDFPage_GetObject(page, k);
-                        if (otherObj && FPDFPageObj_GetType(otherObj) == FPDF_PAGEOBJ_TEXT) {
-                            float otherLeft = 0.0f, otherBottom = 0.0f, otherRight = 0.0f, otherTop = 0.0f;
-                            if (FPDFPageObj_GetBounds(otherObj, &otherLeft, &otherBottom, &otherRight, &otherTop)) {
-                                double otherCenterY = (otherBottom + otherTop) / 2.0;
-                                if (std::abs(otherCenterY - origCenterY) <= tolerance) {
-                                    if (otherLeft >= (origRight - 2.0f)) {
-                                        FPDFPageObj_Transform(otherObj, 1.0, 0.0, 0.0, 1.0, deltaX, 0.0);
-                                        reflowedCount++;
-                                    }
-                                }
-                            }
-                        }
-                    }
-                    spdlog::info("Reflow Engine: shifted {} subsequent text objects on the same line by {}", reflowedCount, deltaX);
-                }
+                bool doSiblingShift = !wasJustified && hasOrigBounds;
 
                 for (int idx : objectIndices) {
                     FPDF_PAGEOBJECT objToRemove = FPDFPage_GetObject(page, idx);
@@ -358,6 +349,7 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
 
                 FPDF_FONT reconFont = nullptr;
                 const fonts::pdf_fonts::ReconstructedFont* reconRf = nullptr;
+                bool reconFullyCovered = false;
                 if (matchedFontInfo && matchedFontInfo->isEmbedded && classifyFontFidelity(*matchedFontInfo) != "exact") {
                     reconRf = lookupReconFont(matchedFontInfo->internalFontId);
                     if (reconRf && reconRf->ok) {
@@ -371,13 +363,9 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                             if (reconFont) loadedFontsCache_[rkey] = reconFont;
                         }
                         if (reconFont) {
-                            bool fullyCovered = true;
+                            reconFullyCovered = true;
                             for (uint32_t cp : unicodeCodepoints)
-                                if (cp >= 0x20 && !reconRf->coveredUnicode.count(cp)) { fullyCovered = false; break; }
-                            if (fullyCovered && !font) {
-                                font = reconFont;
-                                spdlog::info("Tier-2: replace_text exact reconstructed font '{}'", matchedFontInfo->internalFontId);
-                            }
+                                if (cp >= 0x20 && !reconRf->coveredUnicode.count(cp)) { reconFullyCovered = false; break; }
                         }
                     }
                 }
@@ -419,6 +407,11 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                         }
                     }
 
+                    if (!font && reconFont && reconFullyCovered) {
+                        font = reconFont;
+                        spdlog::info("Tier-2: replace_text reconstructed fallback '{}'", matchedFontInfo->internalFontId);
+                    }
+
                     if (!font) {
                         spdlog::info("Font Engine: Loading standard PDF font for replace_text: {}", fontName);
                         font = FPDFText_LoadStandardFont(doc_, fontName.c_str());
@@ -432,7 +425,10 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                         loadedFontsCache_[cacheKey] = font;
                     }
                 }
+
+                if (font && useEmbedded) reconFont = nullptr;
                 
+                std::vector<FPDF_PAGEOBJECT> emittedObjs;
                 if (font) {
                     constexpr unsigned int kRef = 1000;
                     double emToPage = (fontSize > 0.0 ? fontSize : 1.0) * a / static_cast<double>(kRef);
@@ -495,6 +491,7 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                                 FPDFText_SetText(wobj, reinterpret_cast<FPDF_WIDESTRING>(wu.data()));
                                 FPDFPageObj_Transform(wobj, a, b, c, d, penX, f);
                                 FPDFPage_InsertObjectAtIndex(page, wobj, minIndex);
+                                emittedObjs.push_back(wobj);
                             }
                             penX += (wpx[wi] + estSpace) * k + extraPerGap;
                         }
@@ -547,6 +544,7 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                                 FPDFText_SetText(obj, reinterpret_cast<FPDF_WIDESTRING>(seg.data()));
                                 FPDFPageObj_Transform(obj, a, b, c, d, penX, f);
                                 FPDFPage_InsertObjectAtIndex(page, obj, minIndex);
+                                emittedObjs.push_back(obj);
                             }
                             double adv = segWidthPage(segFace, seg8);
                             if (adv <= 0.0) {  // shaping unavailable -> fall back to ink bbox
@@ -588,6 +586,7 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                             FPDFPageObj_Transform(newTextObj, aScale, b, c, d, e, f);
 
                             FPDFPage_InsertObjectAtIndex(page, newTextObj, minIndex);
+                            emittedObjs.push_back(newTextObj);
                         } else {
                             spdlog::error("Failed to create new text object");
                         }
@@ -595,6 +594,28 @@ std::expected<void, EngineError> PdfiumDocument::applyOp_replaceText(const nlohm
                     }
                 }
 
+                if (doSiblingShift) {
+                    double shift = totalWidth - origWidth;
+                    if (std::abs(shift) > 0.05) {
+                        double rowTol = (std::max)(5.0, fontSize * 0.5);
+                        int nObj = FPDFPage_CountObjects(page);
+                        int moved = 0;
+                        for (int k = 0; k < nObj; ++k) {
+                            FPDF_PAGEOBJECT o = FPDFPage_GetObject(page, k);
+                            if (!o || FPDFPageObj_GetType(o) != FPDF_PAGEOBJ_TEXT) continue;
+                            if (std::find(emittedObjs.begin(), emittedObjs.end(), o) != emittedObjs.end()) continue;
+                            float l = 0, bo = 0, rr = 0, tt = 0;
+                            if (!FPDFPageObj_GetBounds(o, &l, &bo, &rr, &tt)) continue;
+                            if (std::abs((bo + tt) / 2.0 - origCenterY) <= rowTol && l >= origRight - 2.0f) {
+                                FPDFPageObj_Transform(o, 1.0, 0.0, 0.0, 1.0, shift, 0.0);
+                                moved++;
+                            }
+                        }
+                        spdlog::info("replace_text: sibling shift {} obj by {:.2f} (deltaX = newW {:.2f} - origW {:.2f})",
+                                     moved, shift, totalWidth, origWidth);
+                    }
+                }
+
                 if (!FPDFPage_GenerateContent(page)) {
                     spdlog::error("Failed to generate page content after replace_text");
                 }

engine/src/parser/pdfium_fonts.cpp

@@ -130,7 +130,7 @@ PdfiumDocument::EmissionFont PdfiumDocument::loadEmissionFont(
         cacheKey += "#" + std::to_string(h);
     }
 
-    if (useEmbedded && reuseFont && !isSubsetFont) {
+    if (useEmbedded && reuseFont) {
         out.font = reuseFont;
         if (auto perObj = getFontDataFromObjects(pageIndex, srcObjects, internalFontId);
             perObj.has_value() && !perObj->empty()) {
@@ -150,35 +150,6 @@ PdfiumDocument::EmissionFont PdfiumDocument::loadEmissionFont(
     }
     if (out.font) return out;
 
-    if (matchedFontInfo && matchedFontInfo->isEmbedded && classifyFontFidelity(*matchedFontInfo) != "exact") {
-        const auto* rf = lookupReconFont(internalFontId);
-        if (rf && rf->ok) {
-            bool covered = true;
-            for (uint32_t cp : codepoints) {
-                if (cp >= 0x20 && !rf->coveredUnicode.count(cp)) { covered = false; break; }
-            }
-            if (covered) {
-                std::lock_guard<std::mutex> lock(loadedFontsMutex_);
-                const std::string rkey = "recon_" + internalFontId;
-                if (loadedFontsCache_.count(rkey)) {
-                    out.font = loadedFontsCache_[rkey];
-                    auto mit = loadedMeasureFaces_.find(rkey);
-                    if (mit != loadedMeasureFaces_.end()) out.measureFace = mit->second;
-                } else {
-                    loadedFontDataBuffers_[rkey] = rf->sfnt;
-                    const auto& bytes = loadedFontDataBuffers_[rkey];
-                    out.font = FPDFText_LoadFont(doc_, bytes.data(), static_cast<uint32_t>(bytes.size()), FPDF_FONT_TRUETYPE, true);
-                    if (out.font) {
-                        loadedFontsCache_[rkey] = out.font;
-                        auto mf = std::make_shared<fonts::FontFace>();
-                        if (mf->loadFromMemory(bytes)) { loadedMeasureFaces_[rkey] = mf; out.measureFace = mf; }
-                    }
-                }
-                if (out.font) { spdlog::info("Tier-2: emit reconstructed embedded font '{}'", internalFontId); return out; }
-            }
-        }
-    }
-
     if (useEmbedded) {
         std::vector<uint8_t> sourceBytes;
         if (auto perObj = getFontDataFromObjects(pageIndex, srcObjects, matchedFontInfo->internalFontId);
@@ -208,6 +179,36 @@ PdfiumDocument::EmissionFont PdfiumDocument::loadEmissionFont(
             }
         }
     }
+    if (!out.font && matchedFontInfo && matchedFontInfo->isEmbedded &&
+        classifyFontFidelity(*matchedFontInfo) != "exact") {
+        const auto* rf = lookupReconFont(internalFontId);
+        if (rf && rf->ok) {
+            bool covered = true;
+            for (uint32_t cp : codepoints) {
+                if (cp >= 0x20 && !rf->coveredUnicode.count(cp)) { covered = false; break; }
+            }
+            if (covered) {
+                std::lock_guard<std::mutex> lock(loadedFontsMutex_);
+                const std::string rkey = "recon_" + internalFontId;
+                if (loadedFontsCache_.count(rkey)) {
+                    out.font = loadedFontsCache_[rkey];
+                    auto mit = loadedMeasureFaces_.find(rkey);
+                    if (mit != loadedMeasureFaces_.end()) out.measureFace = mit->second;
+                } else {
+                    loadedFontDataBuffers_[rkey] = rf->sfnt;
+                    const auto& bytes = loadedFontDataBuffers_[rkey];
+                    out.font = FPDFText_LoadFont(doc_, bytes.data(), static_cast<uint32_t>(bytes.size()), FPDF_FONT_TRUETYPE, true);
+                    if (out.font) {
+                        loadedFontsCache_[rkey] = out.font;
+                        auto mf = std::make_shared<fonts::FontFace>();
+                        if (mf->loadFromMemory(bytes)) { loadedMeasureFaces_[rkey] = mf; out.measureFace = mf; }
+                    }
+                }
+                if (out.font) { spdlog::info("Tier-2: emit reconstructed embedded font '{}' (faithful load unavailable)", internalFontId); return out; }
+            }
+        }
+    }
+
     if (!out.font) {
         out.font = FPDFText_LoadStandardFont(doc_, fontName.c_str());
         if (!out.font) out.font = FPDFText_LoadStandardFont(doc_, "Helvetica");

engine/src/qpdf/qpdf_writer.cpp

@@ -5,10 +5,35 @@
 #include <qpdf/QPDFWriter.hh>
 #include <qpdf/QPDFPageDocumentHelper.hh>
 #include <qpdf/QPDFPageObjectHelper.hh>
+#include <qpdf/QPDFObjectHandle.hh>
+#include <qpdf/Pl_Buffer.hh>
+#include <qpdf/Buffer.hh>
+#include <algorithm>
 #endif
 
 namespace pdfengine::qpdf_layer {
 
+#ifdef PDFENGINE_WITH_QPDF
+namespace {
+class QQDepthCounter : public QPDFObjectHandle::ParserCallbacks {
+public:
+    int depth = 0;
+    int minDepth = 0;
+    void handleObject(QPDFObjectHandle obj, size_t, size_t) override {
+        if (!obj.isOperator()) return;
+        const std::string op = obj.getOperatorValue();
+        if (op == "q") {
+            ++depth;
+        } else if (op == "Q") {
+            --depth;
+            if (depth < minDepth) minDepth = depth;
+        }
+    }
+    void handleEOF() override {}
+};
+}
+#endif
+
 std::expected<void, std::string>
 QpdfWriter::replacePageStreamAndSave(const std::string& sourcePath, 
                                      const std::string& destPath, 
@@ -46,4 +71,98 @@ QpdfWriter::replacePageStreamAndSave(const std::string& sourcePath,
 #endif
 }
 
+std::expected<std::optional<std::vector<uint8_t>>, std::string>
+QpdfWriter::rebalanceContentStreams(const std::vector<uint8_t>& pdfBytes,
+                                    const std::vector<int>& pageIndices) const {
+#ifndef PDFENGINE_WITH_QPDF
+    (void)pdfBytes; (void)pageIndices;
+    return std::unexpected("QPDF support is not enabled in this build.");
+#else
+    try {
+        QPDF pdf;
+        pdf.processMemoryFile("rebalance.pdf",
+                              reinterpret_cast<const char*>(pdfBytes.data()),
+                              pdfBytes.size());
+
+        QPDFPageDocumentHelper pdh(pdf);
+        auto pages = pdh.getAllPages();
+        const int pageCount = static_cast<int>(pages.size());
+
+        std::vector<int> targets = pageIndices;
+        if (targets.empty()) {
+            for (int i = 0; i < pageCount; ++i) targets.push_back(i);
+        }
+
+        bool changed = false;
+        for (int idx : targets) {
+            if (idx < 0 || idx >= pageCount) continue;
+            QPDFPageObjectHelper& page = pages[static_cast<size_t>(idx)];
+
+            QQDepthCounter counter;
+            page.parsePageContents(&counter);
+            if (counter.minDepth >= 0) continue;
+
+            const int n = -counter.minDepth;
+            Pl_Buffer collected("page-contents");
+            page.pipePageContents(&collected);
+            collected.finish();
+            auto raw = collected.getBufferSharedPointer();
+            std::string content(reinterpret_cast<const char*>(raw->getBuffer()), raw->getSize());
+
+            std::string wrapped;
+            for (int k = 0; k < n; ++k) wrapped += "q\n";
+            wrapped.append(content);
+            wrapped.push_back('\n');
+            for (int k = 0; k < n; ++k) wrapped += "Q\n";
+
+            QPDFObjectHandle newStream = QPDFObjectHandle::newStream(&pdf, wrapped);
+            page.getObjectHandle().replaceKey("/Contents", newStream);
+            changed = true;
+        }
+
+        if (!changed) return std::nullopt;
+
+        QPDFWriter writer(pdf);
+        writer.setOutputMemory();
+        writer.setStreamDataMode(qpdf_s_preserve);
+        writer.write();
+        auto out = writer.getBufferSharedPointer();
+        return std::vector<uint8_t>(out->getBuffer(), out->getBuffer() + out->getSize());
+    } catch (const std::exception& e) {
+        return std::unexpected(std::string("QPDF rebalance error: ") + e.what());
+    }
+#endif
+}
+
+std::expected<std::vector<uint8_t>, std::string>
+QpdfWriter::setNeedAppearances(const std::vector<uint8_t>& pdfBytes) const {
+#ifndef PDFENGINE_WITH_QPDF
+    (void)pdfBytes;
+    return std::unexpected("QPDF support is not enabled in this build.");
+#else
+    try {
+        QPDF pdf;
+        pdf.processMemoryFile("appearances.pdf",
+                              reinterpret_cast<const char*>(pdfBytes.data()),
+                              pdfBytes.size());
+
+        QPDFObjectHandle root = pdf.getRoot();
+        if (!root.hasKey("/AcroForm")) return pdfBytes; 
+        QPDFObjectHandle acro = root.getKey("/AcroForm");
+        if (!acro.isDictionary()) return pdfBytes;
+
+        acro.replaceKey("/NeedAppearances", QPDFObjectHandle::newBool(true));
+
+        QPDFWriter writer(pdf);
+        writer.setOutputMemory();
+        writer.setStreamDataMode(qpdf_s_preserve);
+        writer.write();
+        auto out = writer.getBufferSharedPointer();
+        return std::vector<uint8_t>(out->getBuffer(), out->getBuffer() + out->getSize());
+    } catch (const std::exception& e) {
+        return std::unexpected(std::string("QPDF setNeedAppearances error: ") + e.what());
+    }
+#endif
+}
+
 }

engine/src/qpdf/qpdf_writer.hpp

@@ -2,6 +2,8 @@
 
 #include <string>
 #include <vector>
+#include <optional>
+#include <cstdint>
 #include <expected>
 
 class QPDF;
@@ -14,10 +16,19 @@ public:
 
     [[nodiscard]]
     std::expected<void, std::string>
-    replacePageStreamAndSave(const std::string& sourcePath, 
-                             const std::string& destPath, 
-                             int pageIndex, 
+    replacePageStreamAndSave(const std::string& sourcePath,
+                             const std::string& destPath,
+                             int pageIndex,
                              const std::string& newStreamData) const;
+
+    [[nodiscard]]
+    std::expected<std::optional<std::vector<uint8_t>>, std::string>
+    rebalanceContentStreams(const std::vector<uint8_t>& pdfBytes,
+                            const std::vector<int>& pageIndices) const;
+
+    [[nodiscard]]
+    std::expected<std::vector<uint8_t>, std::string>
+    setNeedAppearances(const std::vector<uint8_t>& pdfBytes) const;
 };
 
 }

frontend/src/lib/gatewayService.ts

@@ -262,6 +262,10 @@ export interface ReflowParagraphData {
   lines?: ReflowFragment[][];
   lineBaselineY?: number[];
   lineX?: number[];
+  hangingIndent?: number;
+  listMarker?: string;
+  listKind?: 'bullet' | 'ordered';
+  listLevel?: number;
 }
 
 export interface DeleteAnnotationData {

frontend/src/lib/pdfiumEngine.ts

@@ -18,7 +18,7 @@ function getModule(): Promise<PdfiumModule | null> {
   if (!modulePromise) {
     modulePromise = (async () => {
       try {
-        const V = '20260626-auxfont';
+        const V = '20260630-fontfix3';
         const resp = await fetch(`/pdfium-engine.mjs?v=${V}`, { cache: 'no-store' });
         if (!resp.ok) throw new Error(`pdfium-engine.mjs ${resp.status}`);
         const blobUrl = URL.createObjectURL(new Blob([await resp.text()], { type: 'text/javascript' }));

frontend/src/viewer/ParagraphEditor.tsx

@@ -153,32 +153,117 @@ function resolveStyleEl(node: Text, root: HTMLElement): HTMLElement | null {
   return null;
 }
 
+const BLOCK_TAGS = /^(DIV|P|LI|BLOCKQUOTE|PRE)$/;
+
 function extractFlatRuns(editable: HTMLElement, dominantFid: string, domSize: number, domColor: string): ReflowFragment[] {
   const out: ReflowFragment[] = [];
-  const walker = document.createTreeWalker(editable, NodeFilter.SHOW_TEXT);
-  let node = walker.nextNode() as Text | null;
-  while (node) {
-    const styleEl = resolveStyleEl(node, editable);
-    const fid = styleEl?.getAttribute('data-fid') || dominantFid;
-    const size = parseFloat(styleEl?.getAttribute('data-size') ?? '') || domSize;
-    const color = styleEl?.getAttribute('data-color') ?? domColor;
-    const text = node.textContent ?? '';
-    if (text) {
-      const frag: ReflowFragment = { text, internalFontId: fid, fontSize: size, color };
-      const aRaw = node.parentElement === styleEl ? styleEl?.getAttribute('data-advances') : null;
-      if (aRaw) {
-        try {
-          const a = JSON.parse(aRaw) as number[];
-          if (Array.isArray(a) && a.length === text.length) frag.advances = a;
-        } catch { }
+  const pushBreak = () => {
+    if (!out.length) return;                 // ignore a leading break (no preceding line to end)
+    const prev = out[out.length - 1];
+    if (prev.text === '\n') return;          // collapse consecutive structural breaks
+    out.push({ text: '\n', internalFontId: prev.internalFontId, fontSize: prev.fontSize, color: prev.color });
+  };
+  const walk = (node: Node) => {
+    for (let child = node.firstChild; child; child = child.nextSibling) {
+      if (child.nodeType === Node.TEXT_NODE) {
+        const raw = child.textContent ?? '';
+        if (!raw) continue;
+        const styleEl = resolveStyleEl(child as Text, editable);
+        const fid = styleEl?.getAttribute('data-fid') || dominantFid;
+        const size = parseFloat(styleEl?.getAttribute('data-size') ?? '') || domSize;
+        const color = styleEl?.getAttribute('data-color') ?? domColor;
+        const parts = raw.split('\n');
+        for (let i = 0; i < parts.length; i++) {
+          if (i > 0) pushBreak();
+          const text = parts[i];
+          if (!text) continue;
+          const frag: ReflowFragment = { text, internalFontId: fid, fontSize: size, color };
+          const aRaw = (child as Text).parentElement === styleEl ? styleEl?.getAttribute('data-advances') : null;
+          if (aRaw) {
+            try {
+              const a = JSON.parse(aRaw) as number[];
+              if (Array.isArray(a) && a.length === text.length) frag.advances = a;
+            } catch { }
+          }
+          out.push(frag);
+        }
+      } else if (child.nodeType === Node.ELEMENT_NODE) {
+        const el = child as HTMLElement;
+        if (el.tagName === 'BR') { pushBreak(); continue; }
+        if (BLOCK_TAGS.test(el.tagName)) pushBreak();   // a block boundary is a line break
+        walk(el);
       }
-      out.push(frag);
     }
-    node = walker.nextNode() as Text | null;
-  }
+  };
+  walk(editable);
   return out;
 }
 
+function domTextWithBreaks(editable: HTMLElement): string {
+  let s = '';
+  const walk = (node: Node) => {
+    for (let child = node.firstChild; child; child = child.nextSibling) {
+      if (child.nodeType === Node.TEXT_NODE) s += child.textContent ?? '';
+      else if (child.nodeType === Node.ELEMENT_NODE) {
+        const el = child as HTMLElement;
+        if (el.tagName === 'BR') s += '\n';
+        else { if (BLOCK_TAGS.test(el.tagName) && s && !s.endsWith('\n')) s += '\n'; walk(el); }
+      }
+    }
+  };
+  walk(editable);
+  return s;
+}
+
+function normalizeForCompare(s: string): string {
+  return s.replace(/[ \t]+/g, ' ').replace(/ *\n */g, '\n').replace(/\n+$/, '').replace(/^\n+/, '').trim();
+}
+
+export type ListKind = 'bullet' | 'ordered';
+
+function splitSegmentsByBreak(runs: ReflowFragment[]): ReflowFragment[][] {
+  const segs: ReflowFragment[][] = [[]];
+  for (const r of runs) {
+    const parts = r.text.split('\n');
+    for (let i = 0; i < parts.length; i++) {
+      if (i > 0) segs.push([]);
+      if (parts[i]) segs[segs.length - 1].push({ ...r, text: parts[i] });
+    }
+  }
+  return segs;
+}
+
+let markerMeasureCanvas: HTMLCanvasElement | null = null;
+function measureTextWidth(text: string, sizePt: number, family: string): number {
+  if (!markerMeasureCanvas) markerMeasureCanvas = document.createElement('canvas');
+  const ctx = markerMeasureCanvas.getContext('2d');
+  if (!ctx) return sizePt;                 
+  ctx.font = `${sizePt}px ${family}`;
+  return ctx.measureText(text).width;     
+}
+
+function applyListMarkers(
+  runs: ReflowFragment[],
+  kind: ListKind,
+  dom: { fid: string; size: number; color: string; family: string },
+): { runs: ReflowFragment[]; hangingIndent: number; marker: string } {
+  const segs = splitSegmentsByBreak(runs);
+  let n = 0;
+  let lastMarker = kind === 'ordered' ? '1. ' : '• ';
+  const out: ReflowFragment[] = [];
+  segs.forEach((seg, si) => {
+    if (si > 0) out.push({ text: '\n', internalFontId: dom.fid, fontSize: dom.size, color: dom.color });
+    if (!seg.some((r) => r.text.trim())) { out.push(...seg); return; }
+    n++;
+    const marker = kind === 'ordered' ? `${n}. ` : '• ';
+    lastMarker = marker;
+    out.push({ text: marker, internalFontId: dom.fid, fontSize: dom.size, color: dom.color });
+    out.push(...seg);
+  });
+  const hangingIndent = measureTextWidth(lastMarker, dom.size, dom.family);
+  return { runs: out, hangingIndent, marker: lastMarker };
+}
+
 function globalCaretOffset(el: HTMLElement): number {
   const sel = window.getSelection();
   if (!sel || sel.rangeCount === 0) return 0;
@@ -256,6 +341,9 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
   const [caretBox, setCaretBox] = useState<{ left: number; top: number; height: number } | null>(null);
   const [edited, setEdited] = useState(false);
   const [wasmFailed, setWasmFailed] = useState(false);
+  const [listKind, setListKind] = useState<ListKind | null>(null);
+  const [indentLevel, setIndentLevel] = useState(0);
+  const INDENT_STEP = 18; // pt per nesting level (whole-block indent; per-item nesting is a known limit)
 
   const dominantFid = layout.seedRuns.find((r) => r.text.trim() && r.fid)?.fid
     ?? layout.seedRuns.find((r) => r.fid)?.fid ?? '';
@@ -272,8 +360,22 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
   const editorTop = firstBaselineScreen - (leadingPx + fontPx * 0.7) / 2;
   const bandTop = Math.max(0, Math.min(editorTop - leadingPx * 0.5, firstBaselineScreen - fontPx * 1.15));
 
+  const measureFamily = /times|serif/i.test(domFontName) ? 'Times New Roman, serif'
+    : /courier|mono/i.test(domFontName) ? 'Courier New, monospace' : 'Arial, sans-serif';
+
   const buildReflowData = (runs: ReflowFragment[], origLines?: OrigLine[]) => {
-    const linesData = origLines && origLines.length ? {
+    const listActive = listKind !== null;
+    let outRuns = runs;
+    let effColumnLeft = columnLeft;
+    let listFields: Record<string, unknown> = {};
+    if (listActive) {
+      const { runs: marked, hangingIndent, marker } = applyListMarkers(
+        runs, listKind, { fid: dominantFid, size: domSize, color: domColor, family: measureFamily });
+      outRuns = marked;
+      effColumnLeft = columnLeft + indentLevel * INDENT_STEP;
+      listFields = { hangingIndent, listKind, listLevel: indentLevel, listMarker: marker };
+    }
+    const linesData = (!listActive && origLines && origLines.length) ? {
       lines: origLines.map((l) => l.frags.map((f) => ({
         text: f.text, internalFontId: f.fid, fontSize: f.size, color: f.color,
         ...(f.advances ? { advances: f.advances } : {}),
@@ -283,13 +385,15 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
     } : {};
     return {
       objectIndices: layout.objectIndices,
-      runs: runs.length ? runs : [{ text: ' ', internalFontId: dominantFid, fontSize: domSize, color: '#000000' }],
+      runs: outRuns.length ? outRuns : [{ text: ' ', internalFontId: dominantFid, fontSize: domSize, color: '#000000' }],
       ...linesData,
-      columnLeft, columnRight,
+      columnLeft: effColumnLeft, columnRight,
       pushColumnLeft: pushColumnLeft ?? columnLeft,
       firstBaselineY: layout.firstBaselineY, leading,
-      oldLineCount: layout.oldLineCount, align,
+      oldLineCount: layout.oldLineCount,
+      align: listActive ? 'left' : align,
       paraId: paraIdRef.current,
+      ...listFields,
     };
   };
 
@@ -364,8 +468,6 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
   const renderPreview = async () => {
     const el = editRef.current;
     if (!el) return;
-    // Tier-2 parity: register cmap-augmented embedded fonts with WASM so the reflow preview
-    // reuses the document's real glyphs (matching the native save).
     const fids = new Set<string>([dominantFid, ...layout.seedRuns.map((r) => r.fid)].filter(Boolean));
     await Promise.all([...fids].map((f) => wasmEnsureAuxFont(documentId, f)));
     const dpi = Math.round(72 * zoom);
@@ -452,7 +554,7 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
       el.appendChild(span);
     }
     el.focus();
-    initialTextRef.current = (el.textContent ?? '').replace(/\s+/g, ' ').trim();
+    initialTextRef.current = normalizeForCompare(domTextWithBreaks(el));
   }, []);
 
   const fallbackVisible = wasmFailed && !hasPreview;
@@ -500,8 +602,8 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
     committedRef.current = true;
     const el = editRef.current;
     if (!el) { onCancel(); return; }
-    const nowText = (el.textContent ?? '').replace(/\s+/g, ' ').trim();
-    if (nowText === initialTextRef.current) { onCancel(); return; }
+    const nowText = normalizeForCompare(domTextWithBreaks(el));
+    if (nowText === initialTextRef.current && listKind === null) { onCancel(); return; }
     const flat = extractFlatRuns(el, dominantFid, domSize, domColor);
     if (flat.length === 0) { onCancel(); return; }
     const cv = previewCanvasRef.current;
@@ -520,6 +622,34 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
   };
   const cancel = () => { committedRef.current = true; onOverflowPreview?.([]); onOverflowCaret?.(null); onCancel(); };
 
+  const markEditedAndRender = () => {
+    editedRef.current = true;
+    if (!edited) setEdited(true);
+    positionCaret();
+    scheduleRender();
+  };
+
+  const insertHardBreak = () => {
+    const el = editRef.current;
+    if (!el) return;
+    document.execCommand('insertLineBreak'); // inserts a <br>; extractFlatRuns maps it to '\n'
+    markEditedAndRender();
+  };
+
+  const toggleList = (kind: ListKind) => {
+    setListKind((prev) => (prev === kind ? null : kind));
+    if (listKind === null) setIndentLevel(0);
+    editRef.current?.focus();
+    markEditedAndRender();
+  };
+  const changeIndent = (delta: number) => {
+    setIndentLevel((lv) => Math.max(0, Math.min(5, lv + delta)));
+    editRef.current?.focus();
+    markEditedAndRender();
+  };
+
+  useEffect(() => { if (edited) scheduleRender(); /* eslint-disable-next-line */ }, [listKind, indentLevel]);
+
   return (
     <>
       <style>{`@keyframes pe-caret-blink{0%,49%{opacity:1}50%,100%{opacity:0}}`}</style>
@@ -545,6 +675,36 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
         />
       )}
 
+      {!fallbackVisible && (
+        <div
+          className="absolute z-[40] flex items-center gap-0.5 rounded-md border border-slate-200 bg-white px-1 py-0.5 shadow-md"
+          style={{ left: colLeftPx, top: Math.max(0, editorTop - 34) }}
+          onMouseDown={(e) => e.preventDefault()}
+        >
+          <button
+            type="button"
+            title="Bullet list"
+            onClick={() => toggleList('bullet')}
+            className={`px-1.5 py-0.5 rounded text-sm leading-none ${listKind === 'bullet' ? 'bg-blue-100 text-blue-700' : 'text-slate-600 hover:bg-slate-100'}`}
+          >• List</button>
+          <button
+            type="button"
+            title="Numbered list"
+            onClick={() => toggleList('ordered')}
+            className={`px-1.5 py-0.5 rounded text-sm leading-none ${listKind === 'ordered' ? 'bg-blue-100 text-blue-700' : 'text-slate-600 hover:bg-slate-100'}`}
+          >1. List</button>
+          {listKind !== null && (
+            <>
+              <span className="mx-0.5 h-4 w-px bg-slate-200" />
+              <button type="button" title="Decrease indent (Shift+Tab)" onClick={() => changeIndent(-1)}
+                className="px-1.5 py-0.5 rounded text-sm leading-none text-slate-600 hover:bg-slate-100 disabled:opacity-40" disabled={indentLevel === 0}>⇤</button>
+              <button type="button" title="Increase indent (Tab)" onClick={() => changeIndent(1)}
+                className="px-1.5 py-0.5 rounded text-sm leading-none text-slate-600 hover:bg-slate-100 disabled:opacity-40" disabled={indentLevel >= 5}>⇥</button>
+            </>
+          )}
+        </div>
+      )}
+
       <div
         ref={editRef}
         contentEditable
@@ -561,8 +721,15 @@ export const ParagraphEditor: React.FC<ParagraphEditorProps> = ({
         onKeyUp={positionCaret}
         onKeyDown={(e) => {
           if (e.nativeEvent.isComposing || composingRef.current) return;
-          if (e.key === 'Enter') { e.preventDefault(); commit(); }
-          if (e.key === 'Escape') { e.preventDefault(); cancel(); }
+          if (e.key === 'Enter') {
+            e.preventDefault();
+            if (e.shiftKey) commit();          
+            else insertHardBreak();            
+          } else if (e.key === 'Escape') {
+            e.preventDefault(); cancel();
+          } else if (e.key === 'Tab' && listKind !== null) {
+            e.preventDefault(); changeIndent(e.shiftKey ? -1 : 1);
+          }
         }}
         onBlur={commit}
         onPaste={(e) => { e.preventDefault(); document.execCommand('insertText', false, e.clipboardData.getData('text/plain')); }}

frontend/src/viewer/TextEditLayer.tsx

@@ -52,6 +52,10 @@ export interface ReflowParagraphPayload {
   lines?: ReflowFragment[][];
   lineBaselineY?: number[];
   lineX?: number[];
+  hangingIndent?: number;
+  listMarker?: string;
+  listKind?: 'bullet' | 'ordered';
+  listLevel?: number;
 }
 
 export function isBulletMarker(text?: string): boolean {
@@ -166,7 +170,7 @@ function isFlowingParagraph(para: any): boolean {
 }
 
 function tableCells(line: any): any[] {
-  return (line?.runs ?? []).filter((r: any) => (r.text ?? '').trim());
+  return (line?.runs ?? []).filter((r: any) => (r.text ?? '').trim() && !isBulletMarker(r.text));
 }
 function paraEm(para: any): number {
   let capH = 0;

gateway/app/routers/documents/export.py

@@ -1,7 +1,6 @@
 from fastapi import APIRouter, HTTPException, Response, status
 
 from app.services import engine
-from app.services.export import apply_need_appearances
 from app.services.store import document_store
 
 router = APIRouter(tags=["documents"])
@@ -27,13 +26,11 @@ def export_document(document_id: str):
 
     try:
         doc = d["doc_instance"]
-        bytes_data = doc.save_full()
+        bytes_data = doc.save_full_for_export()
         filename = d["filename"]
         if not filename.endswith(".pdf"):
             filename += ".pdf"
 
-        bytes_data = apply_need_appearances(bytes_data)
-
         return Response(
             content=bytes_data,
             media_type="application/pdf",

gateway/app/routers/edits.py

@@ -246,6 +246,10 @@ class ReflowParagraphData(BaseModel):
     lines: list[list[ReflowRun]] | None = None
     lineBaselineY: list[float] | None = None
     lineX: list[float] | None = None
+    hangingIndent: float = 0.0
+    listMarker: str | None = None
+    listKind: Literal["bullet", "ordered"] | None = None
+    listLevel: int | None = None
 
 
 class ReflowParagraphOperation(BaseModel):

gateway/app/services/export.py

@@ -1,44 +0,0 @@
-"""Export pipeline: full-save + a pypdf pass that sets /NeedAppearances.
-
-NOTE: ``pypdf`` is intentionally imported lazily inside the function. It is not a
-declared gateway dependency and may only be resolvable via the roaming
-site-packages path appended below, so importing it at module load would prevent
-the whole app from starting on machines without it. Behavior here is preserved
-verbatim from the original ``routers.documents.export_document``; if pypdf is
-later added to ``pyproject.toml`` dependencies, the sys.path augmentation can be
-removed and the import hoisted to module scope.
-"""
-
-
-def apply_need_appearances(bytes_data: bytes) -> bytes:
-    """Round-trip the PDF through pypdf to set AcroForm /NeedAppearances=true."""
-    import os
-    import sys
-
-    roaming_path = os.path.join(
-        os.environ.get("APPDATA", "C:\\Users\\azeem\\AppData\\Roaming"),
-        "Python",
-        "Python312",
-        "site-packages",
-    )
-    if roaming_path not in sys.path:
-        sys.path.append(roaming_path)
-
-    import io
-
-    import pypdf
-
-    reader = pypdf.PdfReader(io.BytesIO(bytes_data))
-    writer = pypdf.PdfWriter()
-    writer.append(reader)
-
-    acro_form = writer.root_object.get("/AcroForm")
-    if acro_form is not None:
-        acro_form_dict = acro_form.get_object()
-        acro_form_dict[pypdf.generic.NameObject("/NeedAppearances")] = pypdf.generic.BooleanObject(
-            True
-        )
-
-    out_stream = io.BytesIO()
-    writer.write(out_stream)
-    return out_stream.getvalue()

wasm/CMakeLists.txt

@@ -43,11 +43,15 @@ target_include_directories(pdfengine_wasm PRIVATE
 
 target_link_libraries(pdfengine_wasm PRIVATE pdfengine::pdfengine)
 
+target_link_directories(pdfengine_wasm PRIVATE
+    "${CMAKE_BINARY_DIR}/vcpkg_installed/wasm32-emscripten/lib")
+
 set(_wasm_font_src "${CMAKE_SOURCE_DIR}/engine/assets/fonts")
 set(_wasm_font_dir "${CMAKE_BINARY_DIR}/embed_fonts")
 file(COPY "${_wasm_font_src}/" DESTINATION "${_wasm_font_dir}")
 
 target_link_options(pdfengine_wasm PRIVATE
+    "-Wl,--allow-multiple-definition"
     "--embed-file" "${_wasm_font_dir}@/fonts"
     "-sMODULARIZE=1"
     "-sEXPORT_ES6=1"