[PATCH] D31262: MIParser/MIRPrinter: Compute block successors if not explicitely specified

[Quentin Colombet via llvm-commits]

> On May 1, 2017, at 5:03 PM, Matthias Braun via Phabricator <reviews at reviews.llvm.org> wrote:
> 
> MatzeB added a comment.
> 
> In https://reviews.llvm.org/D31262#742311, @qcolombet wrote:
> 
>> In a nutshell, I am saying that we shouldn't conflict two goals: editing and printing. Printing should be full fledged, editing should be minimal.
> 
> 
> Sounds like we are not in sync what we do with .mir... It is a tool to create llvm lit tests, isn't it?

I agree.

> Printing at least today is mostly done via MachineFunction::print() etc. Sure we want to unify the formats used there and .mir in the long term. But printing with llc -stop-after is the main tool to get started writing a lit test and we should promoted/encourage nice syntax there.

I don’t see why something were everything is explicitly set is not nice. Also things that might be obvious (/computable) from one person may not be so for someone else. Having one full fledge format for printing seems desirable to avoid that.

In the end, given the long term goal is to have .mir and ::print to be the same, why would should we comprise on the output of .mir? It does not bring us any step closer to our end goal. If anything I would prefer a separated strip-mir tool/pass that guys the minimal acceptable input for the .mir parser.

> 
> 
> Repository:
>  rL LLVM
> 
> https://reviews.llvm.org/D31262
> 
> 
>