Latency is an intrinsic property of all human-computer systems. As it can affect user experience and performance, it should be kept as low as possible for real-time applications. To identify the source of latency, measuring partial latencies is necessary. We present a new method for measuring the latency of graphics frameworks on X11-based systems. Our tool measures the time between an input event arriving at the kernel until a pixel is updated in graphics memory. In a systematic evaluation with 36 test applications, we found that our method delivers consistent results for most tested frameworks, and does not add a significant amount of additional end-to-end latency. Even though further investigation is required to explain inconsistencies with Qt-based frameworks, our method measures the latency of graphics frameworks reliably and accurately in all other cases.