Fix Issue 10058 - Inconsistent mangling between C++ and extern(C++).

[ibuclaw]

Fixes http://d.puremagic.com/issues/show_bug.cgi?id=10058

Though maybe the other implementations of ::toCppMangle should be altered to have the same logic as the change here?

Currently it's a half/half situation.

[ibuclaw]

Hold fire on this, am investigating a little more... see the bug report.

[ibuclaw]

Probably should add [DDMD] to this, as I think this might block the ability to have a visitor interface between D and C++ glue, if said visitor interface is defined as extern(C++)

[yebblies]

Add the test cases! static assert(s.mangleof == "...") works for you linux people!

[ibuclaw]

Should probably produce a bunch of test cases for this to make sure that D follows C++ mangling correctly.

[yebblies]

I'd be happy with just the ones from the bug report for now.

Some are available here: https://github.com/mirrors/gcc/tree/master/gcc/testsuite/g++.dg/abi

[ibuclaw]

Didn't see that you commented before me. :)

I'm in London today and without a laptop. Will fix this up either tonight or tomorrow. Most of what I can see in the g++ testsuite are to do with C++ templates...

[ibuclaw]

Waiting for auto-tester.

[yebblies]

Wait - how come this test passes on win32?

[ibuclaw]

I don't think the autotester has ran yet?

Did not check win32. :)

[ghost]

I've edited the title so it's easier to know what the pull is about.

[ibuclaw]

typo in one of the static asserts... serves me right for typing the mangling by hand... :)

[ibuclaw]

ok, linux testers have started passing... did you say that this won't work for windows?

[yebblies]

... Or OSX... cppmangle.c is not even part of the windows build. On OSX you could check the correct OSX mangling is generated... but on windows it's done in the backend and cannot be checked via mangleof. And of course the mangling is different from linux.

[yebblies]

I figure a version(Linux): at the top will fix it for now.

[ibuclaw]

ok, version(linux) is it then...

[WalterBright]

This breaks the following code:

../dmd -m32 -c cppa
g++ -m32 -c cppb.cpp

g++ -m32 cppa.o cppb.o -o cppa -m32 -lphobos2 -lpthread -lm -L../phobos/generated/osx/release/32
Undefined symbols for architecture i386:
  "__Z5foo10PKcS1_", referenced from:
      _D4cppa6test10FZv in cppa.o
  "__Z7foobar9P5elem9S1_", referenced from:
      _D4cppa5test9FZv in cppa.o
ld: symbol(s) not found for architecture i386
collect2: ld returned 1 exit status

--- errorlevel 256

(failure was on OSX)

[IgorStepanov]

@WalterBright please, write contents of cppb.cpp and cppa.d

[yebblies]

@WalterBright Thanks, please put it in bugzilla with a test case (using .mangleof if possible)

[ibuclaw]

@yebblies I'm on it. =)

[ibuclaw]

FYI, regressions:

http://d.puremagic.com/issues/show_bug.cgi?id=4059
http://d.puremagic.com/issues/show_bug.cgi?id=5148