30401 2019-02-23 02:48:32 +0000 [XSLT30] xsl:initial-template must be public, we don't say this, and examples in the spec and tests are wrong 2019-03-12 15:57:46 +0000 1 1 1 Unclassified XPath / XQuery / XSLT XSLT 3.0 Recommendation PC Windows NT NEW P2 normal --- 1 abel.online mike public-qt-comments oldest_to_newest 129553 0 abel.online 2019-02-23 02:48:32 +0000 There are a few issues here: 1) In quite a few tests we use an xsl:package with xsl:initial-template (to name a few, most package-version-xxx). 2) In the spec we have under 3.5.3.5 (reference binding), under "Example: Named Component References in Inline Functions" a package fragment with xsl:initial-template that has no visibility attribute and is therefore private. (there's one other occurrence in a spec-example, but there it appears under xsl:stylesheet, which will default to public). 3) Under 2.3.4 Call-Template Invocation we say that the default for call-template invocation is xsl:initial-template, further in that section we define: [ERR XTDE0040] It is a dynamic error if the invocation of the stylesheet specifies a template name that does not match the expanded QName of a named template defined in the stylesheet, whose visibility is public or final. ---- I vaguely remember we did this on purpose, the reason being that if a library package defines an xsl:initial-template, you don't want that accidentally taking over your starting point, so having it defaulting to private seems reasonable. However, the text is also open to interpretation, the error says: "if the invocation ... *specifies* a template name". You could read that as if you *don't* specify it, the error does not apply (resulting in the default xsl:initial-template to be called regardless of its visibility). But this conflicts with the very reason of this error: it should also be raised if that template does not exist. To sum up: * I think xsl:initial-template ought to have visibility public or final * It seems sensible to allow this to be implicitly true for the any such template in the top-level package, but this may be considered a new feature, which we cannot do at this stage * Tests need to be fixed (I've fixed most locally) For changes to the spec I suggest we choose one of these: * If we cannot change the default (that xsl:initial-template is always valid if it exists with any visibility), then we should add a Note to clarify this * We should fix the example in the spec to give the <xsl:template> there a visibility attribute 129554 1 abel.online 2019-02-23 02:52:36 +0000 Related: Bug #30398 129635 2 abel.online 2019-03-12 15:57:46 +0000 (In reply to Abel Braaksma from comment #0) > * Tests need to be fixed (I've fixed most locally) As can be seen in the referenced bug #30398 most such tests have been fixed and logged under that bug. Leaving this issue open for discussion, and for fixing the relative parts of the spec.