On Wed, Nov 6, 2013 at 12:32 PM, Alp Toker <[email protected]> wrote: > > On 06/11/2013 17:09, Aaron Ballman wrote: >> Ping? >> >> On Fri, Nov 1, 2013 at 5:41 PM, Aaron Ballman <[email protected]> wrote: >>> On Fri, Nov 1, 2013 at 4:15 PM, Reid Kleckner <[email protected]> wrote: >>>> Thanks for tackling and reducing this! This was reported in July, and it's >>>> a shame we didn't report it to them before they went from preview to >>>> release: http://llvm.org/bugs/show_bug.cgi?id=16606 >>> Agreed! >>> >>> And from what I can tell, the patch I provided is probably the best >>> way for us to work around the issue for now; the alternative is to >>> remove the const version of the method from the base class. The good >>> news is, this only seems to affect situations involving this >>> particular pattern, which doesn't appear to be very common in the >>> source base, so we likely are not hitting other miscompiles. >>> >>> ~Aaron > > Confirmed this resolves the problem, and that no other instances of the > bug seem present. > > Aaron, I'd go with whichever of the two has more locality. Is removing > the const version of the method a one-liner?
No, it'd require updating a fair amount of code along with losing some const-correctness. I think the patch as-stands would be the correct workaround. > I think you can go ahead with this and link to the PR in a comment so it > doesn't get "tidied". Thanks for the review (I also confirmed with Reid off-list), and it's been committed in r194176. ~Aaron _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
