[PATCH] D25620: DebugInfo: introduce DIAlignment type

Fri Oct 14 14:07:19 PDT 2016

The important change in this review is going from uint64_t -> uint32_t for the alignment.
This is what we should be discussing here :-)

As for the typedef I think it boils down to bikeshedding:
- Is the typedef giving us more type safety? (looks like the answer is no)
- Is an opaque DIAlignment type any more readable than an explicit uint32_t or unsigned?
Perhaps having a plain "unsigned AlignmentInBits" is just as readable and more in line with the surrounding code.

-- adrian

> On Oct 14, 2016, at 1:54 PM, David Blaikie <dblaikie at gmail.com> wrote:
> 
> (ah, I just assumed from the number of files changed/the description this was more than just adding a typedef - *shrug* I'm less fussed about it, not sure it's worth having a separate typedef for it, but not enough to have any opinion on this code review)
> 
> On Fri, Oct 14, 2016 at 1:51 PM Adrian Prantl <aprantl at apple.com <mailto:aprantl at apple.com>> wrote:
> 
> > On Oct 14, 2016, at 12:02 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com <mailto:dexonsmith at apple.com>> wrote:
> >
> >>
> >> On 2016-Oct-14, at 10:16, David Blaikie <dblaikie at gmail.com <mailto:dblaikie at gmail.com>> wrote:
> >>
> >> Victor - was there any particular benefit/motivation for adding a new type here that you had in mind/had discovered in implementing it? (or precedent you were inspired by - I admit to not having all the DI* hierarchy paged in these days, so perhaps I've missed some obvious prior art)
> >>
> >> On Fri, Oct 14, 2016 at 10:15 AM Adrian Prantl <aprantl at apple.com <mailto:aprantl at apple.com>> wrote:
> >>> On Oct 14, 2016, at 10:13 AM, David Blaikie <dblaikie at gmail.com <mailto:dblaikie at gmail.com>> wrote:
> >>>
> >>>
> >>>
> >>> On Fri, Oct 14, 2016 at 9:59 AM Adrian Prantl <aprantl at apple.com <mailto:aprantl at apple.com>> wrote:
> >>>
> >>>
> >>>> On Oct 14, 2016, at 9:48 AM, David Blaikie <dblaikie at gmail.com <mailto:dblaikie at gmail.com>> wrote:
> >>>
> >>>>
> >>>
> >>>> Could someone point me to where the discussion for adding this type came out of? I didn't spot it at a cursory glance of the previous/existing threads.
> >>>
> >>>>
> >>>
> >>>
> >>>
> >>> This came out of the review thread in:
> >>>
> >>>
> >>>
> >>> D25073: [DebugInfo]: preparation to implement DW_AT_alignment
> >>>
> >>> https://reviews.llvm.org/D25073 <https://reviews.llvm.org/D25073>
> >>>
> >>>
> >>>
> >>> where I argued that we should not be using a full uint64_t for the alignment fields in the DI.* metadata nodes. I don't think the concrete solution of introducing a new alignment type has been discussed here before.
> >>>
> >>> Any particular benefits of introducing a new type? We don't I think have any wrapper types for otherwise singular numeric values, do we? (except maybe DIFlags?)
> >>
> >> I don't have a strong opinion about adding a new type. If we also had a separate type for sizes, we could potentially make the interface more typesafe to avoid accidentally confusing size and alignment, but that's about it.
> >
> > Haven't looked at the patch, so forgive me if I'm confused, but I thought you said this was just a typedef.  In that case, how is it providing any type safety?
> 
> You are right, not as a typedef. We would have to wrap it in a class to benefit from the type safety.
> 
> >>
> >> -- adrian
> >>
> >>>
> >>>
> >>>
> >>>
> >>> -- adrian
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20161014/ce20d5fa/attachment.html>