Work around Nim compiler bug nim-lang/Nim#5140, which breaks

newQSqlDateTime().

qt5_qtsql.nimble

@@ -1,6 +1,6 @@
 [Package]
 name          = "qt5_qtsql"
-version       = "1.0.2"
+version       = "1.0.3"
 author        = "Philip Wernersbach <philip.wernersbach@gmail.com>"
 description   = "Binding for Qt 5's Qt SQL library. Provides a single API for multiple database engines."
 license       = "MIT"

qt5_qtsql/src/qdatetime.nim

@@ -37,15 +37,22 @@ proc newQDateTimeObj*(): QDateTimeObj {.header: QDATETIME_H, importcpp: "QDateTi
 proc setMSecsSinceEpoch*(dateTime: var QDateTimeObj, msecs: qint64) {.header: QDATETIME_H, importcpp: "setMSecsSinceEpoch".}
 proc setTimeSpec*(dateTime: var QDateTimeObj, timeSpec: QtTimeSpec) {.header: QDATETIME_H, importcpp: "setTimeSpec".}
 
-proc newQDateTimeObj*(msecs: qint64): QDateTimeObj =
-    result = newQDateTimeObj()
+# The newQDateTimeObj procs need to be templates to workaround a bug in the Nim compiler.
+# When they are procs, Nim zeros the memory of the "result" variable, which is invalid in
+# C++. (nim-lang/Nim#5140)
+template newQDateTimeObj*(msecs: qint64): QDateTimeObj =
+    var result = newQDateTimeObj()
     result.setMSecsSinceEpoch(msecs)
 
-proc newQDateTimeObj*(msecs: qint64, timeSpec: QtTimeSpec): QDateTimeObj =
-    result = newQDateTimeObj()
+    result
+
+template newQDateTimeObj*(msecs: qint64, timeSpec: QtTimeSpec): QDateTimeObj =
+    var result = newQDateTimeObj()
     result.setTimeSpec(timeSpec)
     result.setMSecsSinceEpoch(msecs)
 
+    result
+
 proc currentQDateTimeUtc*(): QDateTimeObj {.header: QDATETIME_H, importcpp: "QDateTime::currentDateTimeUtc".}
 
 proc toQStringObj*(dateTime: QDateTimeObj, format: cstring): QStringObj {.header: QDATETIME_H, importcpp: "toString".}