[llvm-dev] [LLVM][DISubprogram][LL format updation query] Question regarding moving DISubprogram DIFlags to DISPFlag.

[Robinson, Paul via llvm-dev]

If it makes the patches easier to review, one possibility is to
leave scaffolding in place to support the old textual format while
moving the actual flags around. Then in a follow-up patch, remove
the support for the old textual format and update the .ll tests.
(Of course you would need to preserve support for the old .bc format.)

I think it is reasonable to assume that there are no mixed-mode .ll
files; that is, if you see any of the old bool flags, the file is
using the old format and not the new format.  So you would not need
to worry about merging old-style flags and new-style flags.

--paulr

> -----Original Message-----
> From: llvm-dev <llvm-dev-bounces at lists.llvm.org> On Behalf Of Djordje
> Todorovic via llvm-dev
> Sent: Thursday, February 20, 2020 4:27 AM
> To: Chirag Patel <Chirag at raincode.com>; llvm-dev at lists.llvm.org
> Subject: Re: [llvm-dev] [LLVM][DISubprogram][LL format updation query]
> Question regarding moving DISubprogram DIFlags to DISPFlag.
> 
> I see, since supporting older formats within LLParser is used only to
> avoid updating a bunch of tests (but not required), I do not see any
> obstacle to doing that (as long as we keep the bitcode backward
> compatibility).
> 
> Best,
> Djordje
> 
> On 20.2.20. 10:19, Chirag Patel wrote:
> >> Could you please describe what is the benefit of that?
> > Currently there are two ways to provide DISPFlagDefinition, via bool and
> SPFlag, I would like to make it only via SPFlags, it will be NFC and it
> will make the changes in parser simpler for moving five flags from from
> DIFlags to DISPFlags. Currently parser checks the presence of SPFlags to
> see if the definition is present in bool or spflag if I move flags to
> spflags, it will create problems hence some of the flags may be present in
> spflags and in Boolean, as in example spFlags: DISPFlagThunk, isLocal:
> True.
> >
> > - Chirag.
> >
> >
> > -----Original Message-----
> > From: Djordje Todorovic <djordje.todorovic at rt-rk.com>
> > Sent: 20 February 2020 14:40
> > To: Chirag Patel <Chirag at raincode.com>; llvm-dev at lists.llvm.org
> > Subject: Re: [llvm-dev] [LLVM][DISubprogram][LL format updation query]
> Question regarding moving DISubprogram DIFlags to DISPFlag.
> >
> > Hi Chirag,
> >
> > On 20.2.20. 09:57, Chirag Patel wrote:
> >> Yes, removing the support for isLocal, isDefinition fields completely
> from ll files, currently the LLParser still parses it. I want to remove it
> and update the all the ll files which still uses it.
> >
> > Could you please describe what is the benefit of that?
> >
> >> Also the metadata read will support old format, no changes in that.
> >>
> >> so if ll file has isLocal and isDefinition it will result in parser
> error. But the bitcode read will work as usual.
> >>
> >
> > AFAIK, there are no compatibility guarantees for textual LLVM IR, so we
> only need to support the bitcode auto-upgrade. So, that would be
> acceptable if there is motivation for doing that.
> > In addition, I'm not sure if this was documented somewhere, so we could
> improve the documentation for the metadata backward compatibility.
> >
> >> - Chirag.
> >>
> >> -----Original Message-----
> >> From: Djordje Todorovic <djordje.todorovic at rt-rk.com>
> >> Sent: 20 February 2020 14:16
> >> To: Chirag Patel <Chirag at raincode.com>; llvm-dev at lists.llvm.org
> >> Subject: Re: [llvm-dev] [LLVM][DISubprogram][LL format updation query]
> Question regarding moving DISubprogram DIFlags to DISPFlag.
> >>
> >> Hi Chirag,
> >>
> >> On 20.2.20. 07:51, Chirag Patel via llvm-dev wrote:
> >>> Hello,
> >>>
> >>>
> >>>
> >>> In regard to the review request https://reviews.llvm.org/D74470,
> >>>
> >>> I am trying to move five of the DIFlags to DISPFlag for the moment
> namely DIFlagExplicit, DIFlagPrototyped, DIFlagNoReturn, DIFlagThunk,
> DIFlagAllCallsDescribed.
> >>>
> >>> The llvm ir format for DISubprogram currently has backword
> compatibility where the isLocal, isDefinition, virtuality, isOptimized and
> SPFlags are mutually exclusive.
> >>>
> >>> My question is,
> >>>
> >>> is it a good idea to remove the booleans support'(isLocal,
> isDefinition) and move most of it to spflags and flags in llvm ir?
> >>
> >> But it was already "done", we currently have the 'DISPFlagLocalToUnit'
> and 'DISPFlagDefinition' (please take a look into the
> https://reviews.llvm.org/D54755 and https://reviews.llvm.org/D59288).
> >>
> >> The llvm ir backward compatibility does not list the clear requirements
> on documentations page. This change affects more then 750 ll files.
> >>>
> >>
> >> I am not sure what change will take 750 ll files? Removing the
> 'isLocal' and 'isDefinition'? I think the role of LLVM IR backward
> compatibility is to support interpretation of the old metadata in terms of
> the newest one. Therefore, if there is an 'isDefinition' metadata field,
> that should be interpreted as 'DISPFlagDefinition'.
> >>
> >>
> >>>
> >>> Regards,
> >>>
> >>> Chirag Partel.
> >>>
> >
> > Best,
> > Djordje
> >
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev