[linalg.helpers] too many T

[jwakely]

a subexpression E whose type is T is expression-equivalent to ... includes the declaration template<class T> T abs(T) = delete;

These are not the same T so we should pick a different letter, we have plenty to choose from.

[linalg.helpers.abs]
[linalg.helpers.conj]
[linalg.helpers.real]
[linalg.helpers.imag]

[jwakely]

Reopening, this is about reusing T for different things in wording that was already in the draft, not about the issue in the just-merged [linalg.conj.conjugated] changes.

[tkoeppe]

Oh, sorry! I thought the just-merged changes got rid of the troublesome Ts, but there's more.

[jwakely]

Why does the poison pill for abs use a by-value param and the others use a reference?

template<class T> T abs(T) = delete;
template<class U> U conj(const U&) = delete;
template<class U> U real(const U&) = delete;
template<class U> U imag(const U&) = delete;

I don't think it makes much difference, but if that's the case then there's no reason for them to differ.

[frederick-vs-ja]

I don't think it makes much difference, but if that's the case then there's no reason for them to differ.

Hmm... possibly they catch different combinations of cv-qualification and value category.
However, are they necessary at all (see also P2602R2)? If not, perhaps we can create an LWG issue to remove these poison pills.

[jwakely]

Hmm... possibly they catch different combinations of cv-qualification and value category.

Do they though? I don't think so. All that matters is that the poison pill is found by unqualified lookup, we don't care whether the arguments can be matched to make it viable or not.