<br><br><div class="gmail_quote">On Wed, Sep 21, 2011 at 12:58 PM, David Blaikie <span dir="ltr"><<a href="mailto:dblaikie@gmail.com" target="_blank">dblaikie@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
As per recent discussions on cfe-commits, here's a patch to change all sure-failing assertions to llvm_unreachables (this task was made easier by the canonicalization of these asserts that Richard's been doing with the new boolean conversion from string warning - which meant there were only two major cases ("assert(0 &&" and "assert(false &&" - I don't think I found any message-less asserts in clang though I think there are some in LLVM itself)). The only other thing I had to do was some 80 col wraps due to the increase in length.<div>
<br>If this is the right thing just say the word & I'll check it in.</div>
</blockquote></div><br>(bump)<br><br>Yes/no/maybe? (this'll probably need an update to catch new cases - but I figured I'd check to see if this is preferred before I do that. Perhaps either way we can add something to the documentation to describe when to use each of these devices (assert(false && "foo") or llvm_unreachable("foo")). The same goes for my "default case" patch that I just bumped too - either way, I'll try to document the convention for future reference)<br>
<br>- David<br>