r210584 - Fix crash with x86_64-pc-win32-macho target. <rdar://problem/17235840>

Chandler Carruth chandlerc at google.com
Wed Jun 11 10:45:13 PDT 2014


On Wed, Jun 11, 2014 at 6:38 PM, Bob Wilson <bob.wilson at apple.com> wrote:

> OK, we just have different understanding of “maintaining”.
>
> I agree we do not currently have anyone “maintaining” this platform at the
> level you are looking for. But, there are people relying on it and it
> doesn’t make any sense to threaten to rip it out, just because we don’t
> currently have good tests.
>

No, the reason to rip it out is because it doesn't have active maintainers.
Active maintainers are (IMO) reasonably expected to know enough about the
system to write decent tests for it.

It does sometimes help to threaten to remove features and systems that have
users but aren't being actively maintained, mostly because it usually
causes a maintainer to step forward. We've even does this routinely in the
past with targets. See the sparc backend for a classic example.

There are a lot of things in LLVM for which we do not have good testing. It
> is totally reasonable of you to request that someone add better tests, but
> I don’t know anything more about win32-macho than you do, and I don’t know
> of anyone who is still involved with LLVM who would be in a better position
> to tackle that.
>

If you, a person who is fixing a bug in this system based on user feedback,
neither are a viable maintainer nor know of any viable maintainer, then I
think the only rational conclusion is that it is completely unmaintained.


> But, since I agree with you that more tests would be valuable, I’ve filed
> an internal request for someone to do that:
> <rdar://problem/17271215> add more tests for win32-macho (EFI) platform
>

This is all I was looking for, but if someone doesn't step forward to
maintain this problem actively and reasonably, then just as we do with
targets in LLVM and other platform support, I think that this platform will
need to be removed to keep the project healthy. I don't think there is any
hurry, but eventually the project will have to move forward even if no one
has bothered to maintain this platform and it hurts some users.

-Chandler
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140611/2d815947/attachment.html>


More information about the cfe-commits mailing list