On Wednesday, 30 August 2017 at 21:51:51 UTC, Moritz Maxeiner
wrote:
On Wednesday, 30 August 2017 at 20:23:18 UTC, Johnson Jones
wrote:
It would be nice if, when symbols are missing, they are
unmangled!
Error 42: Symbol Undefined
_D12mMunchhousin12iMunchhousin11__T4GoTsZ4GoMFS12mMunchhousin18__T10MunchhousinTsZ10sMunchhousinfE12mMunchhousin9eGoffZv (void Munchhousin.Munchhousin.Go!(short).Go()
Since that's a linker error and there are a multitude of
linkers that a person could want to use after a D compiler this
is a non-trivial issue in general.
Options to tackle this include:
1) Have a D compiler capture the linker output and demangle it
2) Try to get demangling of D symbols into upstream of the
currently common linkers (GNU linker, gold, lld, etc.)
3) Integrate a (FLOSS) cross platform linker into dmd's backend
at the source code level, with support for such demangling (and
drop OPTLINK)
4) Fork a (FLOSS) cross platform linker for use with D, add
such support, and distribute a binary of it with dmd's binary
distribution (and drop OPTLINK)
I'm not proposing any of these are what should be done, I've
listed them more as an example that something like this would
require extensive discussion.
[1] https://github.com/ldc-developers/ldc/releases/tag/v1.3.0
A simple wrapper around optilink could possible do this? Not
well, but it might be enough?
Obviously optilink is the source of the problem.
How hard is it to take a string like the above and demangle it in
to a readable format? If there is such a funciton in the D
compiler, a simple wrapper could be used to parse the output,
capture symbol errors, and demangle them.
My feeling is that optilink needs to be put to rest. Things die,
let them... move on to greener pastures.
