On 23/11/2008, at 10:03 PM, Joseph Wright wrote: > Will has now added l3quark to the load list for l3tlp, > *but* l3quark already requires l3expan, which itself requires l3tlp. :) Hi Joseph and others, Thanks for tracking this down. I've made the temporary fix as you suggested. I've been doing all my testing with \usepackage{expl3} rather than loading the modules individually. I've been meaning to stop doing this; I've now translated the small amount of code in regression-test that was written in expl3 to plain ol' LaTeX so that module loading itself can be tested by the files in the test suite. (The already- written tests just need a couple of tweaks so the regression results don't need to be changed.) *** This brings up the question: is there any reason for us to keep the modules logically separate? An easy fix might be to add \RequirePackage{expl3} to the top of each l3module. Regarding the best way to fix this particular problem with l3tlp, an alternative is to move \use_none_delimit_by_q_recursion_stop:w into l3basics, where are also defined "\use_none_delimit_by_q_nil:w" and so on. Any thoughts? Will