I discovered yesterday that I had not been thoroughly testing the stylesheets in between each change I made to the "sydb-rub" branch. (Embarrassing mis-communication between me & Martin; my fault, don't ask. :-) So last night I decided to a) write an ant build that would flex them more thoroughly[1] b) run it using the "dev" branch and the "sydb-rub" branch c) IF the output is the same, great, I haven't screwed anything up d) IF the output is different THEN e) run the test on each commit I made to "sydb-rub" to find the error f) FOR EACH error fix it (in the current "sydb-rub") I've done (a), but when I tried (b), I found to my shock that if I 1) switch to "dev" branch 2) run $ saxon odds/odd2odd.xsl Test/base.odd localsource=/path/to/p5subset.xml lang=en 3) I get the error | odds/odd2odd.xsl:487:34: Fatal Error! An empty sequence is not | allowed as the second argument of | fn:resolve-uri() 4) switch to "release" branch 5) run same cmd 6) I get the error | odds/odd2odd.xsl:424: Fatal Error! An empty sequence is not | allowed as the second argument of | resolve-uri() Despite the line number difference (and the weird fact that Saxon can figure out the column number in one case, but not in the other), it is the same error at the same spot in the code: the point in the resolution of <specGrpRef> where we are testing to see if the @target (which is not a local bare name identifier type pointer) is available. The problem (I think) is that the call to base-uri() should be with the argument '/', not the argument "/tei:TEI", as the input document might not actually start with a <TEI> element, it might legally start with a <teiCorpus> element or, as in this case, illegally start with a <schemaSpec>. But what I don't understand is how come our build process hasn't caught this in the past? Can someone try this on their own and double-check that you get the same error? And if so, either someone should dive into our build process and figure out how this was slipping past, or we should cross our fingers until Martin has finished the Test2/ work, and perhaps help him expedite that. Note ---- [1] Since I don't know how to do much in ant, for now the ant build file I wrote just runs odds/odd2odd.xsl on each of the ODD files in Test/.