30405 2019-03-06 12:28:48 +0000 [xslt30] Mode with visibility="private" can still be an eligible initial mode 2019-03-06 16:00:22 +0000 1 1 1 Unclassified XPath / XQuery / XSLT XSLT 3.0 Recommendation PC All NEW P2 normal --- 1 mike mike abel.online public-qt-comments oldest_to_newest 129598 0 mike 2019-03-06 12:28:48 +0000 See test case package-001i Intuitively, the mode is declared with <xsl:mode name="start" visibility="private"/> so one would expect (as the author of the test expects) that the mode is not eligible as an initial mode. However, the mode satisfies rule 5 of the rules for eligibility (§2.3.3): A (named or unnamed) mode M is eligible as an initial mode if one of the following conditions applies, where P is the top-level package of the stylesheet:... (5)The effective value of the declared-modes attribute of the explicit or implicit xsl:package element of P is no, and M appears as a mode-name in the mode attribute of a template rule declared within P. However, §6.6.1 reflects the intuitive expectation: A named mode is not eligible to be used as the initial mode if its visibility is private. So I think rule (5) in §2.3.3 should be amended to read: (5) There is no explicit xsl:mode declaration for mode M, and M appears as a mode-name in the mode attribute of a template rule declared within P. (Note that if there is no explicit xsl:mode declaration, then it can be inferred that the effective value of the declared-modes attribute of the explicit or implicit xsl:package element of P is no) I fear this change will cause a lot of tests to fail, where there is a mode declaration with no visibility attribute. Test streamable-001, for example, has <xsl:mode name="s" streamable="yes"/> and then uses "s" as the initial mode. However, the spec under xsl:mode is fairly clear that (a) visibility="private" is the default, and (b) a mode with private visibility cannot be used as the initial mode. 129602 1 abel.online 2019-03-06 16:00:22 +0000 I'm not sure I fully understand your point here. The title of this bug report suggests that you argue that an xsl:mode with visibility="private" is an eligible initial mode. But from your comments it seems to me you draw a different conclusion, that any named, explicitly private mode cannot be the initial mode (unless it appears in the outermost default-mode attribute). > I fear this change will cause a lot of tests to fail, where there is a mode > declaration with no visibility attribute. Test streamable-001, for example,... True. When I noticed that we often expect the test driver to use a private mode as initial mode, I tried to catch all cases (and clearly missed a few) where this happened and made the initial mode and/or the initial named template public. I originally logged this under bug#30398, and for discussion (as I am not convinced this is necessarily expected behavior), under spec bug#30401.