Patch: Remove unneeded double operation in libstdc++-v3/src/c++17/fs_path.cc

[Jonathan Wakely]

On Fri, 5 Jan 2024 at 13:00, Martin Küttler
<martin.kuettler@kernkonzept.com> wrote:
>
>
> >>This is a small change to libstdc++ which does not change any behavior.
> >
> > Please CC the libstdc++@gcc.gnu.org list on all libstdc++ patches, as
> > documented at https://gcc.gnu.org/lists.html
>
> Acknowledged. Sorry.
>
> >>This change has two, ihmo positive, implications:
> >>
> >> - The implicit conversion from double to int is avoided (Avoiding a
> >>   warning).
> >
> > I don't see any warning here. What do you see?
>
> I see "warning: conversion from ‘double’ to ‘int’ may change value
> [-Wfloat-conversion]" This appears to be a specifically enabled warning.
>
> > Looking at path::_List::reserve now, we probably also want to avoid
> > overflow. Although a path with INT_MAX/1.5 components seems
> > implausible for 32-bit and 64-bit targets, it could be a problem for
> > 16-bit targets. I'll take care of that too.
>
> Nice catch.


We also have some redundant code in path::operator/= which can just be
removed, because _List::reserve does it anyway:

  if (orig_type == _Type::_Multi)
    {
      const int curcap = _M_cmpts._M_impl->capacity();
      if (capacity > curcap)
        capacity = std::max(capacity, (int) (curcap * 1.5));
    }