[llvm-dev] Please dogfood LLD

James Y Knight via llvm-dev llvm-dev at lists.llvm.org
Sun Mar 19 17:15:21 PDT 2017


On Sat, Mar 18, 2017 at 9:06 PM, Sean Silva via llvm-dev <
llvm-dev at lists.llvm.org> wrote:

> I think we should seriously consider adding these hacks (possibly in a way
> that more clearly highlights to users why we had to put them in, such as
> "the following lines are for libtool < vX.Y:"). Realistically, even if we
> land a fix in upstream libtool today, we are unlikely to be able to really
> rely on that in the wild for perhaps 5 years. I don't think our users
> should have to wait that long for LLD to "Just Work" for them. We want LLD
> to have as little "fine print" as possible.
>

I agree. There is really nothing to be gained by forcing everyone to
regenerate their "configure" scripts. Having "lld -v" print something like
"For libtool's benefit: GNU" at the end of the output is a **very** small
price to pay to avoid years of incompatibility with configure scripts which
haven't been generated with a not-even-released-yet version of libtool.

For the second half of the issue, I'd say that having "lld --help" print
what it supports actually seems a sensible thing to do in any case! Might
as well just have it *actually* print out its supported list of targets and
emulations.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170319/61d5c754/attachment.html>


More information about the llvm-dev mailing list