[PATCH] D24091: [Driver] Warn on -nodefaultlibs and -fsanitize

[Chris Bieneman via cfe-commits]

If we do add a warning, we probably need to make the warning more generic than this because it also applies to the Code coverage flags.

-Chris

> On Sep 1, 2016, at 11:29 AM, Bruno Cardoso Lopes <bruno.cardoso at gmail.com> wrote:
> 
> On Wed, Aug 31, 2016 at 11:00 PM, Justin Bogner <mail at justinbogner.com <mailto:mail at justinbogner.com>> wrote:
>> Chris Bieneman <beanz at apple.com> writes:
>>> beanz created this revision.
>>> beanz added reviewers: bogner, zaks.anna, bruno, filcab.
>>> beanz added a subscriber: cfe-commits.
>>> Herald added a subscriber: emaste.
>>> 
>>> The FreeBSD and GNUTools drivers support -fsanitize arguments
>>> bypassing -nodefaultlibs. With https://reviews.llvm.org/D24048, Darwin
>>> will support that behavior as well.
>>> 
>>> To make this a little less magical and behind the curtain this warning
>>> will fire when -nodefaultlibs is used with sanitizer arguments.
>> 
>> I don't think this makes any sense. If it's a legitimate use case to
>> pass these flags together, then anyone who does will have to manually
>> disable this warning - if it's invalid to pass the flags together, we
>> should just disallow it instead of warning. The warning either hurts
>> usability or doesn't do enough.
> 
> I disagree. Assuming that we go for allowing the behavior: I can see
> the fact that the warning may be annoying, but I still think it's
> better for the user to know what's happening and then disable the
> warning, instead of relying on implicit assumptions. Maybe not
> allowing the behavior and better modeling how these things
> interoperate is a better solution, but I'll leave this topic for the
> other thread.
> 
> -- 
> Bruno Cardoso Lopes
> http://www.brunocardoso.cc <http://www.brunocardoso.cc/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20160901/6c9cbc86/attachment-0001.html>