lmi
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lmi] [lmi-commits] master 8ee9eaf 1/7: Don't name unused template p


From: Vadim Zeitlin
Subject: Re: [lmi] [lmi-commits] master 8ee9eaf 1/7: Don't name unused template parameters
Date: Tue, 19 Jan 2021 23:43:18 +0100

On Tue, 19 Jan 2021 20:26:36 +0000 Greg Chicares <gchicares@sbcglobal.net> 
wrote:

GC> let's now consider these:
GC> 
GC> fenv_lmi_x86.hpp: template<typename> friend class control_word;
GC> timer.hpp:        template<typename> friend class AliquotTimer;
GC> 
GC> which are less restrictive. But those do seem appropriate:
GC> 
GC>  - Class 'Timer' is not a template. It befriends AliquotTimer<F>
GC>    for any 'F' we implement.

 Yes, I don't see anything wrong here.

GC>  - The other thing is an x87 control word. We wouldn't change it now,
GC>    even if it were too permissive; and I think the idea was to let
GC>    the orthodox control-word class befriend the msvc one, so it's
GC>    probably correct anyway.

 I'm less sure about this one, I think in any build of lmi only a single
control_word<T> is used and it's, of course, not necessary to be friend
with oneself. I.e. I believe this friend declaration could be just removed
completely. But OTOH I also think we should get rid of this code, and
x87/32 bit version support, completely in the near future so it's hardly a
priority to change it.

 Regards,
VZ

Attachment: pgpdi0Bvrn0kV.pgp
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]