Bugfix 3.3: Import would not report communication errors (#9501)

* arangoimport was neither reporting not stopping on communications errors.  Add reporting of impacted lines for CSV imports.

* correct lines missing from initial backport

CHANGELOG

@@ -1,6 +1,9 @@
 v3.3.24 (XXXX-XX-XX)
 --------------------
 
+* arangoimport would not stop, much less report, communications errors.  Add CSV reporting
+  of line numbers that are impacted during such errors
+
 * changed several internal VelocyPack comparisons in the cluster to not check for bytewise identity,
   but for logical intentity of objects
 

arangosh/Import/ImportFeature.cpp

@@ -452,8 +452,7 @@ void ImportFeature::start() {
 
     std::cout << std::endl;
 
-    // give information about import
-    if (ok) {
+    // give information about import (even if errors occur)
     std::cout << "created:          " << ih.getNumberCreated() << std::endl;
     std::cout << "warnings/errors:  " << ih.getNumberErrors() << std::endl;
     std::cout << "updated/replaced: " << ih.getNumberUpdated() << std::endl;
@@ -463,12 +462,16 @@ void ImportFeature::start() {
       std::cout << "lines read:       " << ih.getReadLines() << std::endl;
     }
 
-    } else {
+
+    if (!ok) {
+      auto const& msgs = ih.getErrorMessages();
+      if (!msgs.empty()) {
         LOG_TOPIC(ERR, arangodb::Logger::FIXME) << "error message(s):";
-      for (std::string const& msg : ih.getErrorMessages()) {
+        for (std::string const& msg : msgs) {
           LOG_TOPIC(ERR, arangodb::Logger::FIXME) << msg;
         }
       }
+    }
   } catch (std::exception const& ex) {
     LOG_TOPIC(ERR, arangodb::Logger::FIXME)
         << "Caught exception " << ex.what() << " during import";

arangosh/Import/ImportHelper.cpp

@@ -815,7 +815,7 @@ void ImportHelper::sendCsvBuffer() {
   SenderThread* t = findIdleSender();
   if (t != nullptr) {
     uint64_t tmp_length = _outputBuffer.length();
-    t->sendData(url, &_outputBuffer);
+    t->sendData(url, &_outputBuffer, _rowOffset +1, _rowsRead);
     addPeriodByteCount(tmp_length + url.length());
   }
 

arangosh/Import/SenderThread.cpp

@@ -47,6 +47,8 @@ SenderThread::SenderThread(std::unique_ptr<httpclient::SimpleHttpClient>&& clien
       _hasError(false),
       _idle(true),
       _ready(false),
+      _lowLineNumber(0),
+      _highLineNumber(0),
       _stats(stats) {}
 
 SenderThread::~SenderThread() {
@@ -62,7 +64,8 @@ void SenderThread::beginShutdown() {
   guard.broadcast();
 }
 
-void SenderThread::sendData(std::string const& url, arangodb::basics::StringBuffer* data) {
+void SenderThread::sendData(std::string const& url, arangodb::basics::StringBuffer* data,
+                            size_t lowLine, size_t highLine) {
   TRI_ASSERT(_idle && !_hasError);
   _url = url;
   _data.swap(data);
@@ -70,12 +73,25 @@ void SenderThread::sendData(std::string const& url, arangodb::basics::StringBuff
   // wake up the thread that may be waiting in run()
   CONDITION_LOCKER(guard, _condition);
   _idle = false;
+  _lowLineNumber = lowLine;
+  _highLineNumber = highLine;
   guard.broadcast();
 }
 
 bool SenderThread::hasError() {
+  bool retFlag = false;
+  {
+    // flag reset after read to prevent multiple reporting
+    //  of errors in ImportHelper
     CONDITION_LOCKER(guard, _condition);
-  return _hasError;
+    retFlag = _hasError;
+    _hasError = false;
+  }
+
+  if (retFlag) {
+    beginShutdown();
+  }
+  return retFlag;
 }
 
 bool SenderThread::isReady() {
@@ -134,6 +150,8 @@ void SenderThread::run() {
 }
 
 void SenderThread::handleResult(httpclient::SimpleHttpResult* result) {
+  bool haveBody = false;
+
   if (result == nullptr) {
     return;
   }
@@ -141,10 +159,12 @@ void SenderThread::handleResult(httpclient::SimpleHttpResult* result) {
   std::shared_ptr<VPackBuilder> parsedBody;
   try {
     parsedBody = result->getBodyVelocyPack();
+    haveBody = true;
   } catch (...) {
-    // No action required
-    return;
+    // no body, likely error situation
   }
+
+  if (haveBody) {
     VPackSlice const body = parsedBody->slice();
 
     // error details
@@ -193,4 +213,16 @@ void SenderThread::handleResult(httpclient::SimpleHttpResult* result) {
       // will trigger the waiting ImportHelper thread to cancel the import
       _hasError = true;
     }
+  } // if
+
+  if (!_hasError && !result->getHttpReturnMessage().empty() && !result->isComplete()) {
+    _errorMessage = result->getHttpReturnMessage();
+    if (0 != _lowLineNumber || 0 != _highLineNumber) {
+      LOG_TOPIC(WARN, arangodb::Logger::FIXME) << "Error left import lines "
+                                                        << _lowLineNumber << " through "
+                                                        << _highLineNumber
+                                                        << " in unknown state";
+    } // if
+    _hasError = true;
+  } // if
 }

arangosh/Import/SenderThread.h

@@ -55,7 +55,8 @@ class SenderThread : public arangodb::Thread {
   /// @brief imports a delimited file
   //////////////////////////////////////////////////////////////////////////////
 
-  void sendData(std::string const& url, basics::StringBuffer* sender);
+  void sendData(std::string const& url, basics::StringBuffer* sender,
+                size_t lowLine = 0, size_t highLine = 0);
 
   bool hasError();
   /// Ready to start sending
@@ -80,6 +81,8 @@ class SenderThread : public arangodb::Thread {
   bool _hasError;
   bool _idle;
   bool _ready;
+  size_t _lowLineNumber;
+  size_t _highLineNumber;
 
   ImportStatistics* _stats;
   std::string _errorMessage;