This is an archive of the discontinued LLVM Phabricator instance.

llvm-config: Add --has-rtti option
ClosedPublic

Authored by tstellarAMD on Aug 7 2015, 2:33 PM.

Details

Summary

This prints NO if LLVM was built with -fno-rtti or an equivalent flag
and YES otherwise. The reasons to add -has-rtti rather than adding -fno-rtti
to --cxxflags are:

  1. Building LLVM with -fno-rtti does not always mean that client

applications need this flag.

  1. Some compilers have a different flag for disabling rtti, and the

compiler being used to build LLVM may not be the compiler being used to
build the application.

Diff Detail

Repository
rL LLVM

Event Timeline

tstellarAMD retitled this revision from to llvm-config: Add --has-rtti option.
tstellarAMD updated this object.
tstellarAMD added a subscriber: llvm-commits.
beanz edited edge metadata.Aug 13 2015, 4:53 PM

Tom,

What about -fno-exceptions? Does we need to handle that flag similarly? There is a thread on llvm-commits today that proposed an alternate solution that covers that case too.

I cc'd you on the thread on llvm-commits, can you please review and provide feedback.

Thanks,
-Chris

Tom,

What about -fno-exceptions? Does we need to handle that flag similarly? There is a thread on llvm-commits today that proposed an alternate solution that covers that case too.

I cc'd you on the thread on llvm-commits, can you please review and provide feedback.

Thanks,
-Chris

Tom,

What about -fno-exceptions? Does we need to handle that flag similarly? There is a thread on llvm-commits today that proposed an alternate solution that covers that case too.

I don't think that compiling LLVM with -fno-exceptions requires library users to use the same, which would mean we don't need a flag for it. However, we should find someone who knows C++ better than I to confirm or refute this.

Tom,

What about -fno-exceptions? Does we need to handle that flag similarly? There is a thread on llvm-commits today that proposed an alternate solution that covers that case too.

I don't think that compiling LLVM with -fno-exceptions requires library users to use the same, which would mean we don't need a flag for it. However, we should find someone who knows C++ better than I to confirm or refute this.

Correct. -fno-exceptions code can link against -fexceptions code. You'll only see issues if you attempt to throw across the boundary (for example in a callback).

jroelofs accepted this revision.Oct 27 2015, 1:33 PM
jroelofs added a reviewer: jroelofs.

LGTM

This revision is now accepted and ready to land.Oct 27 2015, 1:33 PM
This revision was automatically updated to reflect the committed changes.