Establish basic compiler warnings, and fix a few style issues (nomic-…

…ai#3039)

Signed-off-by: Jared Van Bortel <[email protected]>

common/common.cmake

@@ -0,0 +1,44 @@
+function(gpt4all_add_warning_options target)
+    if (MSVC)
+        return()
+    endif()
+    target_compile_options("${target}" PRIVATE
+        # base options
+        -Wall
+        -Wextra
+        # extra options
+        -Wcast-align
+        -Wextra-semi
+        -Wformat=2
+        -Wmissing-include-dirs
+        -Wnull-dereference
+        -Wstrict-overflow=2
+        -Wvla
+        # errors
+        -Werror=format-security
+        -Werror=init-self
+        -Werror=pointer-arith
+        -Werror=undef
+        # disabled warnings
+        -Wno-sign-compare
+        -Wno-unused-parameter
+        -Wno-unused-function
+        -Wno-unused-variable
+    )
+    if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+        target_compile_options("${target}" PRIVATE
+            -Wduplicated-branches
+            -Wduplicated-cond
+            -Wlogical-op
+            -Wno-reorder
+            -Wno-null-dereference
+        )
+    elseif (CMAKE_CXX_COMPILER_ID MATCHES "^(Apple)?Clang$")
+        target_compile_options("${target}" PRIVATE
+            -Wunreachable-code-break
+            -Wunreachable-code-return
+            -Werror=pointer-integer-compare
+            -Wno-reorder-ctor
+        )
+    endif()
+endfunction()

gpt4all-backend/CMakeLists.txt

@@ -1,4 +1,7 @@
 cmake_minimum_required(VERSION 3.23)  # for FILE_SET
+
+include(../common/common.cmake)
+
 set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON)
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
@@ -94,8 +97,6 @@ if (LLMODEL_ROCM)
     list(APPEND BUILD_VARIANTS rocm rocm-avxonly)
 endif()
 
-set(CMAKE_VERBOSE_MAKEFILE ON)
-
 # Go through each build variant
 foreach(BUILD_VARIANT IN LISTS BUILD_VARIANTS)
     # Determine flags
@@ -151,6 +152,7 @@ foreach(BUILD_VARIANT IN LISTS BUILD_VARIANTS)
     # Add each individual implementations
     add_library(llamamodel-mainline-${BUILD_VARIANT} SHARED
         src/llamamodel.cpp src/llmodel_shared.cpp)
+    gpt4all_add_warning_options(llamamodel-mainline-${BUILD_VARIANT})
     target_compile_definitions(llamamodel-mainline-${BUILD_VARIANT} PRIVATE
         LLAMA_VERSIONS=>=3 LLAMA_DATE=999999)
     target_include_directories(llamamodel-mainline-${BUILD_VARIANT} PRIVATE
@@ -169,6 +171,7 @@ add_library(llmodel
     src/llmodel_c.cpp
     src/llmodel_shared.cpp
 )
+gpt4all_add_warning_options(llmodel)
 target_sources(llmodel PUBLIC
     FILE_SET public_headers TYPE HEADERS BASE_DIRS include
     FILES include/gpt4all-backend/llmodel.h

gpt4all-backend/include/gpt4all-backend/llmodel.h

@@ -146,7 +146,7 @@ class LLModel {
     virtual bool supportsEmbedding() const = 0;
     virtual bool supportsCompletion() const = 0;
     virtual bool loadModel(const std::string &modelPath, int n_ctx, int ngl) = 0;
-    virtual bool isModelBlacklisted(const std::string &modelPath) const { (void)modelPath; return false; };
+    virtual bool isModelBlacklisted(const std::string &modelPath) const { (void)modelPath; return false; }
     virtual bool isEmbeddingModel(const std::string &modelPath) const { (void)modelPath; return false; }
     virtual bool isModelLoaded() const = 0;
     virtual size_t requiredMem(const std::string &modelPath, int n_ctx, int ngl) = 0;

gpt4all-backend/src/llmodel_shared.cpp

@@ -260,7 +260,7 @@ void LLModel::generateResponse(std::function<bool(int32_t, const std::string&)>
         cachedTokens.push_back(new_tok.value());
         cachedResponse += new_piece;
 
-        auto accept = [this, &promptCtx, &cachedTokens, &new_tok, allowContextShift]() -> bool {
+        auto accept = [this, &promptCtx, &new_tok, allowContextShift]() -> bool {
             // Shift context if out of space
             if (promptCtx.n_past >= promptCtx.n_ctx) {
                 (void)allowContextShift;

gpt4all-chat/CMakeLists.txt

@@ -1,5 +1,7 @@
 cmake_minimum_required(VERSION 3.25)  # for try_compile SOURCE_FROM_VAR
 
+include(../common/common.cmake)
+
 set(APP_VERSION_MAJOR 3)
 set(APP_VERSION_MINOR 4)
 set(APP_VERSION_PATCH 1)
@@ -157,6 +159,7 @@ qt_add_executable(chat
     src/xlsxtomd.cpp              src/xlsxtomd.h
     ${CHAT_EXE_RESOURCES}
 )
+gpt4all_add_warning_options(chat)
 
 qt_add_qml_module(chat
     URI gpt4all

gpt4all-chat/src/chat.h

@@ -33,7 +33,7 @@ class Chat : public QObject
     Q_PROPERTY(ResponseState responseState READ responseState NOTIFY responseStateChanged)
     Q_PROPERTY(QList<QString> collectionList READ collectionList NOTIFY collectionListChanged)
     Q_PROPERTY(QString modelLoadingError READ modelLoadingError NOTIFY modelLoadingErrorChanged)
-    Q_PROPERTY(QString tokenSpeed READ tokenSpeed NOTIFY tokenSpeedChanged);
+    Q_PROPERTY(QString tokenSpeed READ tokenSpeed NOTIFY tokenSpeedChanged)
     Q_PROPERTY(QString deviceBackend READ deviceBackend NOTIFY loadedModelInfoChanged)
     Q_PROPERTY(QString device READ device NOTIFY loadedModelInfoChanged)
     Q_PROPERTY(QString fallbackReason READ fallbackReason NOTIFY loadedModelInfoChanged)

gpt4all-chat/src/chatllm.cpp

@@ -585,7 +585,7 @@ bool ChatLLM::loadNewModel(const ModelInfo &modelInfo, QVariantMap &modelLoadPro
 
     modelLoadProps.insert("$duration", modelLoadTimer.elapsed() / 1000.);
     return true;
-};
+}
 
 bool ChatLLM::isModelLoaded() const
 {

gpt4all-chat/src/chatmodel.h

@@ -65,8 +65,8 @@ struct ChatItem
     Q_PROPERTY(bool thumbsDownState MEMBER thumbsDownState)
     Q_PROPERTY(QList<ResultInfo> sources MEMBER sources)
     Q_PROPERTY(QList<ResultInfo> consolidatedSources MEMBER consolidatedSources)
-    Q_PROPERTY(QList<PromptAttachment> promptAttachments MEMBER promptAttachments);
-    Q_PROPERTY(QString promptPlusAttachments READ promptPlusAttachments);
+    Q_PROPERTY(QList<PromptAttachment> promptAttachments MEMBER promptAttachments)
+    Q_PROPERTY(QString promptPlusAttachments READ promptPlusAttachments)
 
 public:
     QString promptPlusAttachments() const

gpt4all-chat/src/database.cpp

@@ -296,10 +296,12 @@ static bool selectAllUncompletedChunks(QSqlQuery &q, QHash<IncompleteChunk, QStr
     while (q.next()) {
         QString collection = q.value(0).toString();
         IncompleteChunk ic {
-            /*embedding_model =*/ q.value(1).toString(),
-            /*chunk_id        =*/ q.value(2).toInt(),
-            /*folder_id       =*/ q.value(3).toInt(),
-            /*text            =*/ q.value(4).toString(),
+            /*EmbeddingKey*/ {
+                .embedding_model = q.value(1).toString(),
+                .chunk_id        = q.value(2).toInt(),
+            },
+            /*folder_id =*/ q.value(3).toInt(),
+            /*text      =*/ q.value(4).toString(),
         };
         chunks[ic] << collection;
     }
@@ -1659,7 +1661,7 @@ void Database::scanQueue()
         if (info.isPdf()) {
             QPdfDocument doc;
             if (doc.load(document_path) != QPdfDocument::Error::None) {
-                qWarning() << "ERROR: Could not load pdf" << document_id << document_path;;
+                qWarning() << "ERROR: Could not load pdf" << document_id << document_path;
                 return updateFolderToIndex(folder_id, countForFolder);
             }
             title    = doc.metaData(QPdfDocument::MetaDataField::Title).toString();

gpt4all-chat/src/database.h

@@ -176,7 +176,6 @@ class ChunkStreamer {
     QString                                m_author;
     QString                                m_subject;
     QString                                m_keywords;
-    bool                                   m_atStart;
 
     // working state
     QString                                m_chunk; // has a trailing space for convenience

gpt4all-chat/src/modellist.cpp

@@ -504,7 +504,7 @@ ModelList::ModelList()
     connect(MySettings::globalInstance(), &MySettings::contextLengthChanged, this, &ModelList::updateDataForSettings);
     connect(MySettings::globalInstance(), &MySettings::gpuLayersChanged, this, &ModelList::updateDataForSettings);
     connect(MySettings::globalInstance(), &MySettings::repeatPenaltyChanged, this, &ModelList::updateDataForSettings);
-    connect(MySettings::globalInstance(), &MySettings::repeatPenaltyTokensChanged, this, &ModelList::updateDataForSettings);;
+    connect(MySettings::globalInstance(), &MySettings::repeatPenaltyTokensChanged, this, &ModelList::updateDataForSettings);
     connect(MySettings::globalInstance(), &MySettings::promptTemplateChanged, this, &ModelList::updateDataForSettings);
     connect(MySettings::globalInstance(), &MySettings::systemPromptChanged, this, &ModelList::updateDataForSettings);
     connect(&m_networkManager, &QNetworkAccessManager::sslErrors, this, &ModelList::handleSslErrors);
@@ -520,12 +520,12 @@ QString ModelList::compatibleModelNameHash(QUrl baseUrl, QString modelName) {
     QCryptographicHash sha256(QCryptographicHash::Sha256);
     sha256.addData((baseUrl.toString() + "_" + modelName).toUtf8());
     return sha256.result().toHex();
-};
+}
 
 QString ModelList::compatibleModelFilename(QUrl baseUrl, QString modelName) {
     QString hash(compatibleModelNameHash(baseUrl, modelName));
     return QString(u"gpt4all-%1-capi.rmodel"_s).arg(hash);
-};
+}
 
 bool ModelList::eventFilter(QObject *obj, QEvent *ev)
 {
@@ -2103,7 +2103,7 @@ void ModelList::parseDiscoveryJsonFile(const QByteArray &jsonData)
     emit discoverProgressChanged();
     if (!m_discoverNumberOfResults) {
         m_discoverInProgress = false;
-        emit discoverInProgressChanged();;
+        emit discoverInProgressChanged();
     }
 }
 
@@ -2180,7 +2180,7 @@ void ModelList::handleDiscoveryItemFinished()
 
     if (discoverProgress() >= 1.0) {
         m_discoverInProgress = false;
-        emit discoverInProgressChanged();;
+        emit discoverInProgressChanged();
     }
 
     reply->deleteLater();

gpt4all-chat/src/mysettings.cpp

@@ -186,7 +186,7 @@ void MySettings::restoreModelDefaults(const ModelInfo &info)
     setModelTemperature(info, info.m_temperature);
     setModelTopP(info, info.m_topP);
     setModelMinP(info, info.m_minP);
-    setModelTopK(info, info.m_topK);;
+    setModelTopK(info, info.m_topK);
     setModelMaxLength(info, info.m_maxLength);
     setModelPromptBatchSize(info, info.m_promptBatchSize);
     setModelContextLength(info, info.m_contextLength);