Fix MP4 probe when moov exceeds the window

kernel/include/eac3joc_core.h

@@ -68,6 +68,15 @@ EJOC_API ejoc_renderer_handle EJOC_CALL ejoc_renderer_create(void);
 EJOC_API void EJOC_CALL ejoc_renderer_destroy(ejoc_renderer_handle handle);
 EJOC_API int EJOC_CALL ejoc_renderer_reset(ejoc_renderer_handle handle);
 EJOC_API int EJOC_CALL ejoc_renderer_set_threads(ejoc_renderer_handle handle, uint32_t total_threads);
+/*
+Enables or disables the Ls/Rs band-0 21-tap DC compensation. The caller derives
+it from the JOC downmix configuration: only configurations 3 and 4 enable the
+filter. When disabled, band 0 keeps the common per-band processing (surround
+delay plus -j rotation) instead of being overwritten by the FIR. The delay line
+and DC history advance either way, so the flag may change between frames.
+Defaults to enabled when never called.
+*/
+EJOC_API int EJOC_CALL ejoc_renderer_set_dc_filter(ejoc_renderer_handle handle, uint32_t enabled);
 EJOC_API uint32_t EJOC_CALL ejoc_renderer_thread_count(ejoc_renderer_handle handle);
 EJOC_API const char* EJOC_CALL ejoc_renderer_last_error(ejoc_renderer_handle handle);
 

kernel/src/joc_core/eac3joc_core.cpp

@@ -125,6 +125,10 @@ public:
         return error_[0] ? error_ : "";
     }
 
+    void set_dc_filter(const bool enabled) noexcept {
+        dc_filter_enabled_ = enabled;
+    }
+
     int process(
         const float* bed5,
         const float* lfe,
@@ -305,6 +309,7 @@ private:
                 for (int i = 0; i < 4; ++i) {
                     dc_buffer[20 + i] = current[i][0];
                 }
+                if (dc_filter_enabled_) {
                     for (int slot = 0; slot < 4; ++slot) {
                         Complex sum{0.0, 0.0};
                         for (int tap = 0; tap < 21; ++tap) {
@@ -316,6 +321,7 @@ private:
                         }
                         x_[channel][0][group + slot] = {2.0 * sum.re, 2.0 * sum.im};
                     }
+                }
                 for (int i = 0; i < 20; ++i) {
                     surround_history_[surround][i] = dc_buffer[i + 4];
                 }
@@ -641,6 +647,8 @@ private:
     float analysis_phase_;
     alignas(64) Complex surround_delay_[2][10][64];
     alignas(64) Complex surround_history_[2][20];
+    // band-0 的 21-tap DC 补偿开关；仅 downmix 配置 3/4 由调用方置位。
+    bool dc_filter_enabled_ = true;
     alignas(64) double lfe_delay_[kLfeDelay];
     alignas(64) double matrix_previous_[15][5][64];
     alignas(64) double synthesis_state_[15][640];
@@ -696,6 +704,14 @@ int EJOC_CALL ejoc_renderer_set_threads(ejoc_renderer_handle handle, uint32_t to
     return static_cast<ejoc::Renderer*>(handle)->set_threads(total_threads);
 }
 
+int EJOC_CALL ejoc_renderer_set_dc_filter(ejoc_renderer_handle handle, uint32_t enabled) {
+    if (!handle) {
+        return -1;
+    }
+    static_cast<ejoc::Renderer*>(handle)->set_dc_filter(enabled != 0);
+    return 0;
+}
+
 uint32_t EJOC_CALL ejoc_renderer_thread_count(ejoc_renderer_handle handle) {
     if (!handle) {
         return 0;

kernel/src/joc_core/objects16.cpp

@@ -52,6 +52,15 @@ Status rebuild_objects16(ejoc_renderer_handle handle, const joc_frame_params& pa
     }
 
     out16->assign(static_cast<std::size_t>(JOC_OUTPUT_CHANNELS) * JOC_FRAME_SAMPLES, 0.0f);
+    // band-0 的 21-tap DC 补偿只在 downmix 配置 3/4 下启用，其余配置 band 0
+    // 走与其他 band 相同的处理。
+    const bool dc_filter = params.dmx_config_idx == 3 || params.dmx_config_idx == 4;
+    if (ejoc_renderer_set_dc_filter(handle, dc_filter ? 1u : 0u) != 0) {
+        if (error != nullptr) {
+            *error = "ejoc_renderer_set_dc_filter failed";
+        }
+        return Status::fail(JOC_ERR_RENDER_FAILED, stage::kDsp, "ejoc_renderer_set_dc_filter failed");
+    }
     const int result = ejoc_renderer_process(
         handle, bed5_planar, lfe, params.present_mask, n_bands, n_dpoints, slope_idx, offset_ts,
         dq.data(), params.clipgain, 0.0625f, gain, out16->data());

src/container_scan.cpp

@@ -200,15 +200,24 @@ Result scan_mp4(const Window& file, std::size_t max_bytes) {
     Result result;
     result.kind = Kind::kMp4;
 
-    // moov is usually at the start for streamed files and at the end otherwise.
+    // moov is usually at the start for streamed files and at the end otherwise.  The
+    // walk reads box headers where they lie and steps over mdat in one go, so the
+    // whole file is walked from the top: a fixed prefix window cannot reach a moov
+    // that is larger than the window, which is what an MP4 with its cover art stored
+    // as a video track produces (a 4.9 MB moov against a 4 MiB window).  The tail
+    // range stays as a fallback for a file whose leading boxes do not parse.
     struct Range {
         std::uint64_t begin;
         std::uint64_t end;
     };
-    std::vector<Range> ranges{{0, (std::min<std::uint64_t>)(file.size(), max_bytes)}};
+    std::vector<Range> ranges{{0, file.size()}};
     if (file.size() > max_bytes) {
         ranges.push_back({file.size() - max_bytes, file.size()});
     }
+    // An oversized moov is a file this probe cannot describe, and walking it would be
+    // unbounded work; it is reported instead of entered.
+    constexpr std::uint64_t kMaxMoovBytes = 64ull * 1024ull * 1024ull;
+    bool oversized_moov = false;
 
     unsigned audio_seen = 0;
     bool found_any_audio = false;
@@ -218,6 +227,10 @@ Result scan_mp4(const Window& file, std::size_t max_bytes) {
                    [&](const std::string& type, std::uint64_t payload, std::uint64_t box_end,
                        std::size_t) {
                        if (type != "moov") return true;
+                       if (box_end - payload > kMaxMoovBytes) {
+                           oversized_moov = true;
+                           return true;
+                       }
                        walk_boxes(file, payload, box_end, 1,
                                   [&](const std::string& inner, std::uint64_t inner_payload,
                                       std::uint64_t inner_end, std::size_t) {
@@ -273,6 +286,8 @@ Result scan_mp4(const Window& file, std::size_t max_bytes) {
         result.detail = "mp4: first audio track is " +
                         (first_audio_format.empty() ? std::string("unknown")
                                                     : first_audio_format);
+    } else if (oversized_moov) {
+        result.detail = "mp4: moov is larger than the scan limit";
     } else {
         result.detail = "mp4: no audio track found in the scanned window";
     }

src/joc_decode.cpp

@@ -30,6 +30,32 @@ constexpr std::size_t kBedReadBytes = 48u * 1024u;
 constexpr std::size_t kBedPipeBytes = 1u << 20;
 constexpr std::size_t kFrameSamples = JOC_FRAME_SAMPLES;
 
+// The decoded bed has to stay on the E-AC-3 frame grid that the JOC matrix and the
+// OAMD are indexed by.  An mp4/mov edit list trims the decoded audio instead, which
+// takes the bed off that grid by however much the list removes -- a Dolby Atmos
+// download loses 2432 samples (1.58 frames) -- and every frame's matrix would then be
+// applied to audio tens of milliseconds away from it, so a subset of the objects comes
+// out attenuated.  The demuxer's own option drops the trim; the option exists only on
+// the mov/mp4 demuxer, so it is passed for that family alone.
+bool has_mov_timeline(const std::string& path) {
+    static const char* const kExtensions[] = {".mp4", ".m4a", ".m4b", ".m4v",
+                                              ".mov", ".3gp", ".3g2", ".mj2"};
+    std::string lowered = path;
+    for (char& character : lowered) {
+        if (character >= 'A' && character <= 'Z') {
+            character = static_cast<char>(character - 'A' + 'a');
+        }
+    }
+    for (const char* extension : kExtensions) {
+        const std::size_t length = std::strlen(extension);
+        if (lowered.size() >= length &&
+            lowered.compare(lowered.size() - length, length, extension) == 0) {
+            return true;
+        }
+    }
+    return false;
+}
+
 // ---------------------------------------------------------------------------
 // Kernel entry points.
 //
@@ -685,6 +711,7 @@ bool Engine::Impl::start_bed(std::uint64_t source_sample, std::string* error) {
     // at the start of the file, which is a low-level, noise-like difference from a
     // play-through rather than a misalignment: the position stays exact.
     std::wstring input_arguments = L"-drc_scale 0 -target_level 0";
+    if (has_mov_timeline(input_path)) input_arguments += L" -ignore_editlist 1";
     const std::wstring offset = seek_time(source_sample);
     if (!offset.empty()) input_arguments += L" -ss " + offset;
     std::wstring bed_arguments = L"-map 0:a:";

src/main.cpp

@@ -12,7 +12,7 @@
 
 // Kept in one place: the string reported to foobar2000 and written to the log
 // must not drift apart.
-#define JOC_VERSION "0.2.0"
+#define JOC_VERSION "0.2.2"
 
 DECLARE_COMPONENT_VERSION("JOC decoder (E-AC-3 JOC)", JOC_VERSION,
                           "Plays E-AC-3 JOC (Dolby Atmos) files: the JOC objects are "

src/prefs.cpp

@@ -189,6 +189,7 @@ private:
                     }
                     if (pick_file(m_hwnd, path, title, filter)) {
                         set_text(GetDlgItem(m_hwnd, IDC_EDIT_HRTF), path);
+                        update_status();
                         notify();
                     }
                     return TRUE;
@@ -219,6 +220,10 @@ private:
                     return TRUE;
                 }
                 if (code == EN_CHANGE || code == CBN_SELCHANGE) {
+                    if (id == IDC_EDIT_HRTF || id == IDC_COMBO_LAYOUT ||
+                        id == IDC_EDIT_GAIN) {
+                        update_status();
+                    }
                     notify();
                     return TRUE;
                 }
@@ -338,35 +343,36 @@ private:
     void update_enabled_state() { update_status(); }
 
     void update_status() {
-        const joc_settings::Values values = joc_settings::read();
-        std::string status;
+        const joc_settings::Values values = read_controls();
+        std::string output_status;
         if (values.output != 0) {
             char text[96] = {};
             const unsigned channels = joc_decode::speaker_channels(values.speaker_layout);
             std::snprintf(text, sizeof(text), "扬声器布局 %s（%u 声道）",
                           values.speaker_layout.c_str(), channels);
-            status = text;
-        } else {
-            // Show what will actually be read, so an empty box is not a mystery.
+            output_status = text;
+        }
+        set_text(GetDlgItem(m_hwnd, IDC_LABEL_OUTPUT_STATUS), output_status);
+
         joc_decode::Settings effective = joc_settings::current();
         effective.hrtf_source =
             static_cast<joc_decode::HrtfSource>(values.hrtf_source);
         effective.hrtf_file = values.hrtf_file;
         const std::string file = joc_decode::resolve_hrtf_file(effective);
-            status = std::string(joc_settings::hrtf_source_name(values.hrtf_source)) + "：" +
-                     (file.empty() ? std::string("无法确定默认路径")
-                                   : (values.hrtf_file.empty() ? "默认 " + file : file));
-        }
-        // Say what is actually in effect, including the parts that do not apply
-        // to the selected output mode, so nothing has to be greyed out.
+        std::string hrtf_status =
+            std::string(joc_settings::hrtf_source_name(values.hrtf_source)) + "：";
+        if (values.hrtf_file.empty()) hrtf_status += "默认";
+        hrtf_status += "\n" +
+                       (file.empty() ? std::string("无法确定默认路径") : file);
+        set_text(GetDlgItem(m_hwnd, IDC_LABEL_STATUS), hrtf_status);
+
         char gain[96] = {};
         if (values.gain_enabled) {
-            std::snprintf(gain, sizeof(gain), "\n增益开 %.2f dB", values.gain_db);
+            std::snprintf(gain, sizeof(gain), "增益开 %.2f dB", values.gain_db);
         } else {
-            std::snprintf(gain, sizeof(gain), "\n增益关（输出不衰减）");
+            std::snprintf(gain, sizeof(gain), "增益关（输出不衰减）");
         }
-        status += gain;
-        set_text(GetDlgItem(m_hwnd, IDC_LABEL_STATUS), status);
+        set_text(GetDlgItem(m_hwnd, IDC_LABEL_GAIN_STATUS), gain);
     }
 
     bool changed() const {

src/prefs.rc

@@ -7,11 +7,11 @@
 // sample page uses (SDK\foobar2000\foo_sample\foo_sample.rc).  A CAPTION statement
 // implies WS_CAPTION, which would draw a second frame inside the host's and shift
 // every control down by the caption height -- visible but not clickable.
-IDD_JOC_PREFS DIALOGEX 0, 0, 330, 226
+IDD_JOC_PREFS DIALOGEX 0, 0, 330, 273
 STYLE DS_SETFONT | WS_CHILD
 FONT 8, "Microsoft Sans Serif", 400, 0, 0x0
 BEGIN
-    GROUPBOX        "输出", IDC_GROUP_OUTPUT, 7, 5, 316, 64
+    GROUPBOX        "输出", IDC_GROUP_OUTPUT, 7, 5, 316, 61
     CONTROL         "双耳（HRTF）", IDC_RADIO_BINAURAL, "Button",
                     BS_AUTORADIOBUTTON | WS_GROUP | WS_TABSTOP, 16, 20, 72, 10
     CONTROL         "扬声器布局", IDC_RADIO_SPEAKER, "Button",
@@ -22,31 +22,33 @@ BEGIN
     LTEXT           "双耳模式", IDC_LABEL_MODE, 172, 22, 42, 10
     COMBOBOX        IDC_COMBO_MODE, 216, 19, 96, 90,
                     CBS_DROPDOWNLIST | WS_VSCROLL | WS_TABSTOP
-    LTEXT           "房间尾音（秒）", IDC_LABEL_TAIL, 172, 41, 46, 10
-    EDITTEXT        IDC_EDIT_TAIL, 220, 38, 40, 13, ES_AUTOHSCROLL
+    LTEXT           "房间尾音（秒）", IDC_LABEL_TAIL, 172, 41, 62, 10
+    EDITTEXT        IDC_EDIT_TAIL, 238, 38, 40, 13, ES_AUTOHSCROLL
+    LTEXT           "", IDC_LABEL_OUTPUT_STATUS, 16, 52, 300, 10
 
-    GROUPBOX        "HRTF", IDC_GROUP_HRTF, 7, 74, 316, 84
+    GROUPBOX        "HRTF", IDC_GROUP_HRTF, 7, 71, 316, 100
     CONTROL         "SOFA 文件", IDC_RADIO_HRTF_SOFA, "Button",
-                    BS_AUTORADIOBUTTON | WS_GROUP | WS_TABSTOP, 16, 89, 56, 10
+                    BS_AUTORADIOBUTTON | WS_GROUP | WS_TABSTOP, 16, 86, 56, 10
     CONTROL         "Rosella 个性化模型", IDC_RADIO_HRTF_ROSELLA, "Button",
-                    BS_AUTORADIOBUTTON | WS_TABSTOP, 78, 89, 78, 10
-    LTEXT           "HRTF 文件（留空＝默认位置）", IDC_LABEL_HRTF, 16, 106, 92, 10
-    EDITTEXT        IDC_EDIT_HRTF, 110, 103, 164, 13, ES_AUTOHSCROLL
-    PUSHBUTTON      "浏览…", IDC_BROWSE_HRTF, 279, 103, 38, 13
+                    BS_AUTORADIOBUTTON | WS_TABSTOP, 78, 86, 78, 10
+    LTEXT           "HRTF 文件（留空＝默认位置）", IDC_LABEL_HRTF, 16, 105, 106, 10
+    EDITTEXT        IDC_EDIT_HRTF, 126, 102, 148, 13, ES_AUTOHSCROLL
+    PUSHBUTTON      "浏览…", IDC_BROWSE_HRTF, 279, 102, 38, 13
     LTEXT           "留空时读组件目录下 HRTF\\binaural.sofa 或 binaural.personalized_headphone；SOFA 由渲染器内部编译，滤波组表已编入库中。",
-                    IDC_LABEL_HRTF_HINT, 16, 122, 300, 30
+                    IDC_LABEL_HRTF_HINT, 16, 121, 300, 20
+    LTEXT           "", IDC_LABEL_STATUS, 16, 145, 300, 20
 
-    GROUPBOX        "增益", IDC_GROUP_GAIN, 7, 162, 152, 56
+    GROUPBOX        "增益", IDC_GROUP_GAIN, 7, 176, 316, 48
     CONTROL         "应用增益", IDC_CHECK_GAIN, "Button",
-                    BS_AUTOCHECKBOX | WS_TABSTOP, 16, 178, 58, 10
-    LTEXT           "dB", IDC_LABEL_GAIN, 80, 178, 12, 10
-    EDITTEXT        IDC_EDIT_GAIN, 96, 175, 40, 13, ES_AUTOHSCROLL
+                    BS_AUTOCHECKBOX | WS_TABSTOP, 16, 192, 58, 10
+    LTEXT           "dB", IDC_LABEL_GAIN, 80, 192, 12, 10
+    EDITTEXT        IDC_EDIT_GAIN, 96, 189, 40, 13, ES_AUTOHSCROLL
+    LTEXT           "", IDC_LABEL_GAIN_STATUS, 145, 192, 171, 10
     LTEXT           "双耳渲染后可能超过 0 dBFS，可用负值衰减。",
-                    IDC_LABEL_GAIN_HINT, 16, 194, 134, 20
+                    IDC_LABEL_GAIN_HINT, 16, 208, 300, 10
 
-    GROUPBOX        "其它", IDC_GROUP_MISC, 165, 162, 158, 56
-    LTEXT           "ffmpeg.exe", IDC_LABEL_FFMPEG, 173, 178, 44, 10
-    EDITTEXT        IDC_EDIT_FFMPEG, 216, 175, 62, 13, ES_AUTOHSCROLL
-    PUSHBUTTON      "浏览…", IDC_BROWSE_FFMPEG, 282, 175, 35, 13
-    LTEXT           "", IDC_LABEL_STATUS, 173, 194, 144, 20
+    GROUPBOX        "其它", IDC_GROUP_MISC, 7, 229, 316, 38
+    LTEXT           "ffmpeg.exe", IDC_LABEL_FFMPEG, 16, 245, 44, 10
+    EDITTEXT        IDC_EDIT_FFMPEG, 64, 242, 210, 13, ES_AUTOHSCROLL
+    PUSHBUTTON      "浏览…", IDC_BROWSE_FFMPEG, 279, 242, 38, 13
 END

src/resource.h

@@ -12,6 +12,7 @@
 #define IDC_COMBO_MODE 2016
 #define IDC_LABEL_TAIL 2017
 #define IDC_EDIT_TAIL 2018
+#define IDC_LABEL_OUTPUT_STATUS 2019
 
 #define IDC_GROUP_HRTF 2020
 #define IDC_RADIO_HRTF_SOFA 2021
@@ -26,6 +27,7 @@
 #define IDC_LABEL_GAIN 2032
 #define IDC_EDIT_GAIN 2033
 #define IDC_LABEL_GAIN_HINT 2034
+#define IDC_LABEL_GAIN_STATUS 2035
 
 #define IDC_GROUP_MISC 2040
 #define IDC_LABEL_FFMPEG 2041