For MSVC no exceptions builds do not use _CPPUNWIND

[DBJDBJ]

martinmoene/expected-lite#27 (comment)

Just a friendly "single place" reminder for all nonstd projects

[martinmoene]

Replace

# if defined(__cpp_exceptions) || defined(__EXCEPTIONS) || defined(_CPPUNWIND)

with

# if _MSC_VER
#  include <vcruntime.h>  // for _HAS_EXCEPTIONS
# endif
# if defined(__cpp_exceptions) || defined(__EXCEPTIONS) || (_HAS_EXCEPTIONS)

Plan:

• Update releases below:
• any-lite -xxx
• atomic-lite (being developed)
• bit-lite (NA)
• boolean-lite (NA)
• byte-lite (NA)
• expected-lite - xxx/v0.6.0
• invoke-lite (NA)
• observer-ptr-lite (NA)
• optional-bare - xxx
• optional-fun-lite (NA)
• optional-lite - xxx/v3.4.0
• ring-span-lite (NA)
• scope-lite (NA)
• span-lite - xxx/v0.9.0
• status-value-lite - xxx
• string-view-lite - xxx/v1.7.0
• type-lite (NA)
• value-ptr-lite - xxx
• variant-lite - xxx/v2.1.0

[DBJDBJ]

Thanks, @martinmoene , great :)

Over here is my unfinished result on the subject of MS STL and /kernel.

Kind regards

[martinmoene]

Thanks

Will need some time to further the exceptional behaviour. Likely want to make it configurable under Windows to either use assert() or SEH.

Kind regards

[DBJDBJ]

Good, you made that decision,

SEH is intrinsic to Windows, thus your future customers, Windows teams using that /kernel switch will very likely expect nonstd to raise one or few Structured Exception.

Thanks ...

ps: since my readme.md over there is atrocious. I need to simplify it. I don't understand it myself.

[DBJDBJ]

Will need some time to further the exceptional behavior. Likely want to make it configurable under Windows to either use assert() or SEH.

Well, the "fail fast" movement is certainly raising ... And any of the three (compilers) is here to oblige ...

ps: the side issue -- is this not a great real-life example of why sometimes single include file policy fails :) I mean no disrespect but now you need to travel around repositories and change things, instead of having one common header (single place) where one could change this kind of top-level issues. I am going back to the "single top-level set of headers" model. Ugh.