Unbreak libbacktrace symbolizer

[jakubjelinek]

Summary

FillModuleInfo now has a new argument, but the libbacktrace symbolizer has not been adjusted for that.
This patch fixes that. Additionally I've included a fix for maybe uninitialized warning in ubsan, where an unknown CFITypeCheckKind value
would fall through to the reporting.

[kcc]

LGTM
Do you need me to commit it?
(If you don't have commit access yet, I 'm pretty sure you qualify for one long ago)

Also: is it worth adding a build mode that uses libbacktrace to the LLVM build system (and bots)?
This way we'll avoid breaking this code in future.

[jakubjelinek]

In D38991#899950, @kcc wrote:

LGTM
Do you need me to commit it?

If you could, I'd appreciate it.

(If you don't have commit access yet, I 'm pretty sure you qualify for one long ago)

Also: is it worth adding a build mode that uses libbacktrace to the LLVM build system (and bots)?
This way we'll avoid breaking this code in future.

Maybe, but it would involve adding libbacktrace and the libbacktrace wrappers needed to use it somewhere, probably more work than to check it once a year during merge into GCC.

[kcc]

lib/ubsan/ubsan_handlers_cxx.cc

126

clang doesn't like this:

llvm/projects/compiler-rt/lib/ubsan/ubsan_handlers_cxx.cc:126:3: warning: default label in switch which covers all enumeration values [-Wcovered-switch-default]
  default:
  ^

Do you need this part?