[llvm] r250222 - Tweak to r250117 and change to use ErrorOr and drop isSizeValid for

[Kevin Enderby via llvm-commits]

Hi Rafael,

I have been working on this the last couple of days with help form Lang Hames.  It as been a bit messy as even the iterators had to change.  Still not there yet but the email below is where Lang and I are at the moment.  Hope to have the rest of the changes for the tree soon.

Kev

> On Oct 16, 2015, at 1:20 PM, Lang Hames <lhames at apple.com> wrote:
> 
> Hi Pete, Kevin,
> 
> I decided to try tackling this by making child_iterator contain an ErrorOr<Child> instead of a Child. Partial patch attached. There’s still a bunch of failures in the llvm/tools/* directory because now we have to error check the iterators every time we increment, but the core infrastructure for the change is now in place.
> 
> Cheers,
> Lang.
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Child_ErrorOr.patch
Type: application/octet-stream
Size: 28619 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151016/2ef2e7eb/attachment.obj>
-------------- next part --------------


> On Oct 16, 2015, at 1:14 PM, Rafael Espíndola <rafael.espindola at gmail.com> wrote:
> 
> On 13 October 2015 at 17:23, Rafael Espíndola
> <rafael.espindola at gmail.com> wrote:
>>> I didn’t know how far you wanted me to push this up the stack.  I can do that.  But there are lots of errors that are not handled well in libObject.
>> 
>> Not handling an error is a bug.
>> 
>> Ideally we want to push them up until we get out of lib, at which
>> point we are in tools and can handle the error in whatever way is
>> appropriate for that tool.
>> 
>> But we can use report_fatal_error to move the error checking out one
>> step at a time, but please never eat an error.
> 
> Ping. We are still eating the error and the test is still reverted.
> 
> Cheers,
> Rafael