If you have N cases, for each ith case in (1,2,...,N), you can use the rest data (except ith case) to build a classifier model, then apply this model on the ith case to get its class. After repeat this procedure N times, all cases will be assigned a class label and you can evaluate the accuracy of your classification model.
You randomly devide your data set into k sub samples, then use k-1 sub samples as training data and one as testing data. You repeat this procedure k times.
In addition to responses of Abdelmalek and Wen, suppose you have n observations. You randomly select one observation (as test observation). Build a discriminant function (either linear or quadratic) based on (n-1) observations and the remaining one (test observation) is used to validating the discriminant analysis, then compute the error based on number of misclassification (cross validation error). You may repeat the process and compute average error.