MSC /utf-8 vs /execution-charset:utf-8

[torsten48]

fmt CMakeLists.txt with option FMT_UNICODE=ON enforces /utf-8 to allow UNICODE (UTF8) using MSC/clang-cl on Windows.
/utf-8 is a combination of two switches, /execution-charset:utf-8 and /source-charset:utf-8.
fmt works very well with just /execution-charset:utf-8 and coding utf literals via "\uXX" sequences (and set of FMT_UNICODE = OFF).
With /utf-8 (and as a result /source-charset:utf-8) the usage of utf8 encoded source files are forced.
There is imho no need to enforce utf8 source file encoding.
I would like to replace /utf-8 by /execution-charset:utf-8

Personal background: I'm still using a very special folding editor (Andys Editor, ae), unfortunately not compatible with codepage 65002 (utf8), so /source-charset:utf-8 is a pain for me.

Addition:
The fmt code does not care about FMT_UNICODE, maybe the whole cmake option is obsolete. The compiler itself complains if /execution-charset:utf-8 is not set and utf "\uXX" literals are used, and it also complains if unescaped utf8 is found in the sources while /source-charset:utf-8 is not specified.

[vitaut]

This is not quite correct. {fmt} very much cares about it and in general having different source and literal encoding may lead to data loss. It has nothing to do with codepage 65001, {fmt} supports Unicode on Windows regardless of the active codepage (ACP) or console codepage.

[torsten48]

There is only ONE place where FMT_UNICODE is used, in fmt/base.h where you check if the literal encoding is utf8, exactly what /execution-charset:utf-8 is used for.

/source-charset:utf-8 is not a must and the compiler itself complains if there is utf8 in source code and /source-charset:utf-8 is not specified.

It is absolutely no problem save your header and cpp in ANSI while using utf8 literals.
However, I am apparently a lone dinosaur who does not yet store his source files in utf8.
And I can workaround with disabling FMT_UNICODE and nevertheless using utf8 literals.

[vitaut]

{fmt} uses UTF-8 both in headers and in tests so setting source encoding to UTF-8 is correct and we shouldn't change this. It also affects user code. Disabling FMT_UNICODE to support legacy encoding in your case is the way to go.