[Bug gdb/31331] Wenum-constexpr-conversion should be fixed, soon treated as a hard error

["tromey at sourceware dot org" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=31331

--- Comment #30 from Tom Tromey <tromey at sourceware dot org> ---
IIUC the current state is:

- there's no way to detect whether an enum has an explicit
  underlying type
- we can't use the current approach to detecting if the underlying
  type is signed
- we can't force an underlying type to be unsigned because it
  would break the compile plugin ABI
- but we can't rely on the underlying type of C-like enums to
  be unsigned since that is compiler dependent
- we can't use the promotion detection since that is too broad

It seems to me that there's no need to support a signed
enumeration constant in enum-flags.  It just isn't useful.

So, could all the work be done in an unsigned type that is
big enough, and then if the enum is signed, mask off the
high bit (and assert that no constant with the high bit
set is ever passed in)?

For enums with a signed underlying type this would mean
sacrificing one flag value.  But that seems ok.

For unsigned enums it would all stay the same.

operator~ could just be defined unconditionally.

-- 
You are receiving this mail because:
You are on the CC list for the bug.