[PATCH] D45639: [Driver] Support default libc++ library location on Darwin

Fri Apr 23 09:00:35 PDT 2021

JDevlieghere added a comment.

In D45639#2706605 <https://reviews.llvm.org/D45639#2706605>, @ldionne wrote:

> In D45639#2706589 <https://reviews.llvm.org/D45639#2706589>, @dexonsmith wrote:
>
>> I'm not sure I'm totally following, but just want to double check that the tests won't somehow use the libc++ from the SDK instead of ToT? I guess the test uses `-nostdinc++` or something?
>>
>> Assuming that's doing what we want... I also wanted to highlight this adds a dependency on bot's host clang working with the ToT libc++ headers. That's probably fine -- I assume we keep the bot's Xcode relatively up-to-date -- but I know @ldionne at some point was trying to avoid requiring new libc++ to work with older clangs.

The simulator tests are now using both headers and library from the SDK.

> LLDB should either use the libc++ headers from the SDK *and* the dylib from the SDK, or build libc++ from scratch and test against both the trunk headers and the trunk dylib built for their target, not for the host as they seem to do right now. I'm following offline with @JDevlieghere to try and disentangle that.

As discussed on Slack on Wednesday, lldb is doing the former for the simulator tests. All other tests are building for the host and are using the just-built headers and library.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D45639/new/

https://reviews.llvm.org/D45639