Test bugs in libraries

I didn't think of this myself, but overheard it in a question by someone at a presentation of Yehuda Katz, but I thought it was a nice tip:

If you work around a bug in a library you use, write a test to see if the bug is happening.

That way, once you update the library, and the bug is fixed, your tests will fail, and you know you can now remove the workaround code (of course add comments to your test indicating where workarounds are located).