D4/D5 destructor and other manglings not specified about cxx-abi HOT 5 OPEN

Sorry for letting this sit for a long time. Technically, everything not explicitly set aside for vendor extensions is reserved for future use by the ABI. This is the first I've heard of these manglings, but I would appreciate if vendors asked for these manglings to be dedicated to these purposes when needed.

@zygoloid @jicama

from cxx-abi.

Note that the ABI does already reserve a name for a concept, the "complete object allocating constructor", that it does not actually specify rules for and is definitely not a mandatory part of the ABI. (It is presumably meant to have the effect of return new T(args...) and might be a useful code-size optimization despite not being mandatory in the ABI.)

from cxx-abi.

The CI mangling is covered by issue #48.

Do we need to document C4/D4? (Are they still in use?)

C5/D5 were proposed to the ABI list here: https://www.mail-archive.com/[email protected]/msg00168.html (complete with a patch).

from cxx-abi.

Ah, I knew I should've searched the active issues. Opened an issue for C5/D5 here: #92

from cxx-abi.

Judging from today's gcc code, C/D4 are still used, unlike C/D5.
58a644cfdee53275e3d07eff0c2126dc88b87aef: libcc1/libcp1plugin.cc: plugin_define_cdtor_clone
Even if vendor-specific, just mentioning that such existed at some point in time in a HISTORIC section or so would be quite useful - if only to avoid reusing C/D4.

from cxx-abi.