Essentially the complement test involves crossing true-breeding mutants to see whether or not their mutations are related. Two genes complement each other if one set of genes compensates for another such that the offspring heterozygous for both will revert to the wild type phenotype. If two mutations don't complement, it may mean that they are on the same gene. For example you have two mutant plants. The wild type has blue flowers. The two mutants you have are white. If you cross them, the offspring is blue. That means that the two mutations are likely on separate genes, and that in the cross, one parent gives a normal wt allele to mask the other parent's mutant phenotype on one gene, and the other parent gives a wt allele on a different gene which masks the mutant phenotype of the first parent. Say, for instance one mutation causes an inability to manufacture the blue pigmentation, the other mutation inhibits the deposition of the pigment onto the flower. The hybrid offspring has one working allele for each of the two genes, so it expresses the blue flower phenotype
However, in a similar hypothetical situation, if the heterozygous offspring cross still remained white, it may be that the two mutations in each parent were actually in the same gene, in different locations such that the offspring inherits two crap alleles for a gene responsible for the blue coloration. This is not an definite conclusion. The offspring could also be white because the mutations are on different genes, but the presence of one mutant allele of one gene immediately precludes the effect of the other. (e.g. in the first example, the mutation from one parent instead causes the breakdown of the blue pigmentation, which precludes the color-deposition gene from doing anything at all, and the offspring still has white flowers.).
