Convert "must"s in normative wording that do not indicate logical #1747
Conversation
|
Check that whitespace! |
source/templates.tex
Outdated
| @@ -583,18 +583,17 @@ | |||
| id-expression | |||
| \end{bnf} | |||
|
|
|||
| \begin{note} | |||
There was a problem hiding this comment.
Is this change relevant to the subject at hand?
There was a problem hiding this comment.
I think so. The below "name must be known to refer to a template" is supposed to be what normatively states what is and is not a template-name, but is hopelessly vague, and changing the "must" to a "shall" is also wrong -- the right thing is to say that an identifier is a template-name if name lookup associates it with one, which is what the note prior to the broken sentence says.
That said, I'd be happy to take a less-invasive alternate path through here.
There was a problem hiding this comment.
Let's just leave this one alone for C++17. I'm now concerned that this sentence may be intended to have normative impact (eg, what else would make sizeof &operator""_foo < 8 > 0 parse as a comparison rather than as having a template-argument-list?)
source/templates.tex
Outdated
| instantiation. In the case of a non-type template argument, | ||
| instantiation. | ||
| \begin{note} | ||
| In the case of a non-type template argument, |
There was a problem hiding this comment.
We've discussed this recently for core issue 2307. Why is it safe to turn the normative text into a note here?
There was a problem hiding this comment.
N is not equivalent to, say, (N) per [temp.over.link]p5. But this wording is not particularly clear on whether [temp.over.link] is what we mean by "equivalent". Hmm.
I'm happy to make a different change if you'd prefer. Perhaps something more in this direction:
A template argument that is equivalent to a template parameter (i.e., has the same type as the template parameter or, for a non-type template argument, is an expression that names the template parameter but not as a subexpression) can be used in place of that template parameter in a reference to the current instantiation.
... but that's pretty awkward. Or perhaps:
A template argument that is equivalent to a template parameter (i.e., has the same constant value or the same type as the template parameter) can be used in place of that template parameter in a reference to the current instantiation. In the case of a non-type template argument, this only applies if the argument was given the value of the template parameter and not an expression in which the template parameter appears as a subexpression.
There was a problem hiding this comment.
Well, we don't really mean the "equivalent" from [temp.over.link] here: the example claims that my_I and I are "equivalent", but they are not the same per the ODR (i.e. token-by-token identical). So, we need the normative text to explain the different notion of "equivalent". [The very fact we're even discussing this makes me uneasy about such a large-ish last-minute change.]
That said, your second proposal above seems to be least invasive, so I'd prefer that.
necessity to a different form. Normative requirements use "shall". Advice to users uses "should". Description of the behavior of the implementation that does not of itself constitute a requirement uses the imperative mood. Fixes ISO 21 (C++17 DIS)
necessity to a different form.
Normative requirements use "shall". Advice to users uses "should".
Description of the behavior of the implementation that does not of
itself constitute a requirement uses the imperative mood.
Fixes ISO 21 (C++17 DIS)