diff --git a/poincare/src/integral.cpp b/poincare/src/integral.cpp
index 15fbb9e7e..404ca9902 100644
--- a/poincare/src/integral.cpp
+++ b/poincare/src/integral.cpp
@@ -26,10 +26,19 @@ Expression * Integral::cloneWithDifferentOperands(Expression** newOperands,
 }
 
 float Integral::approximate(Context& context) const {
-  /* We here use Gauss-Legendre quadrature with n = 5 */
+  /* We here use Gauss-Legendre quadrature with n = 5
+   * Gauss-Legendre abscissae and weights taken from
+   * http://www.holoborodko.com/pavel/numerical-methods/numerical-integration/*/
+  /* TODO: implement an adaptive quadrature version
+   * Use Gauss-Konrad quadrature (the one used in GNU scientific library)?
+   * QAGS and QAGI from netlib?
+   * Add to this version:
+   * - assess error
+   * - higher degree if error > threshold
+   * - find a way to reuse x and w ? Or useless, store all values ? */
   XContext xContext = XContext(&context);
-  static float x[5]={0.1488743389f, 0.4333953941f, 0.6794095682f, 0.8650633666f, 0.9739065285f};
-  static float w[5]={0.2955242247f, 0.2692667193f, 0.2190863625f, 0.1494513491f, 0.0666713443f};
+  static float x[5]={0.1488743389816312108848260f, 0.4333953941292471907992659f, 0.6794095682990244062343274f, 0.8650633666889845107320967f, 0.9739065285171717200779640f};
+  static float w[5]={0.2955242247147528701738930f, 0.2692667193099963550912269f, 0.2190863625159820439955349f, 0.1494513491505805931457763f, 0.0666713443086881375935688f};
   float a = m_args[1]->approximate(context);
   float b = m_args[2]->approximate(context);
   float xm = 0.5f*(a+b);