[cfe-dev] [RFC] Upstreaming proposed change to .dia format from Swift

[Jan Korous via cfe-dev]

Hi Owen,

I think this is a great idea!

I'm just thinking about how to make sure the URLs in diagnostics are valid and stay that way.

We might want to use a combination of CI jobs and phabricator checks to make sure all the URLs we produce in diagnostics in both current master and somewhat recent releases are ok.


Jan

> On May 19, 2020, at 9:14 AM, Owen Voorhees via cfe-dev <cfe-dev at lists.llvm.org> wrote:
> 
> Thanks for pointing me towards those docs, they seem like excellent candidates for linking from diagnostics! Providing more information when encountering errors related to things like attributes and command line args that users may not encounter on a regular basis is exactly in line with my goals when adding this.
> 
> It seems apparent that I probably shouldn’t assume documentation will be installed locally when generalizing this feature to work with Clang. Swift ships the docs alongside the compiler primarily because the language and diagnostics are still evolving rapidly, and we don’t have a good mechanism to host versioned documentation at the moment. On the other hand, linking to something like https://clang.llvm.org/docs/LanguageExtensions.html#non-standard-c-11-attributes <https://clang.llvm.org/docs/LanguageExtensions.html#non-standard-c-11-attributes> from Clang diagnostics seems perfectly sensible, and there’s no reason I can think of not to support both web and file URLs.
> 
> I’ll look into Clang’s diagnostics infrastructure to get a better idea of how documentation links can be integrated throughout before moving ahead with the serialization format updates. I think my originally proposed changes are still a little too Swift-specific.
> 
> — Owen
> 
>> On May 19, 2020, at 7:49 AM, Finkel, Hal J. <hfinkel at anl.gov <mailto:hfinkel at anl.gov>> wrote:
>> 
>> 
>> From: Aaron Ballman <aaron.ballman at gmail.com <mailto:aaron.ballman at gmail.com>>
>> Sent: Tuesday, May 19, 2020 7:44 AM
>> To: Finkel, Hal J. <hfinkel at anl.gov <mailto:hfinkel at anl.gov>>
>> Cc: cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org> <cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org>>; Owen Voorhees <owenvoorhees at gmail.com <mailto:owenvoorhees at gmail.com>>
>> Subject: Re: [cfe-dev] [RFC] Upstreaming proposed change to .dia format from Swift
>>  
>> On Tue, May 19, 2020 at 6:26 AM Finkel, Hal J. <hfinkel at anl.gov <mailto:hfinkel at anl.gov>> wrote:
>> >
>> > Hi, Owen,
>> >
>> > I think having documentation links associated with diagnostics is a useful feature, and I would be happy to see it in Clang.
>> 
>> I agree, I think this is a great idea! Hopefully we can also make use
>> of it (or something like it) in clang-tidy as well.
>> 
>> > I would encourage the addition of appropriate links to Clang's https://clang.llvm.org/docs/LanguageExtensions.html <https://clang.llvm.org/docs/LanguageExtensions.html> and, moreover, maybe we can generate these for mis-applied attributes in some automated way (given that our attribute docs are also TableGen-generated).
>> 
>> Are you envisioning something like `[[gnu::nonnull]] int *i;` giving
>> you a diagnostic about the attribute being ignored because it doesn't
>> apply to local variables, along with a link directly to the
>> documentation for that attribute? If so, that's a neat idea! We could
>> possibly even do something like that for unknown or misspelled command
>> line arguments, because the command line reference is also table
>> generated.
>> 
>> 
>> Yes, exactly. I'm glad that you agree 🙂
>> 
>>  -Hal
>> 
>> 
>> ~Aaron
>> 
>> >
>> >  -Hal
>> >
>> > Hal Finkel
>> > Lead, Compiler Technology and Programming Languages
>> > Leadership Computing Facility
>> > Argonne National Laboratory
>> >
>> > ________________________________
>> > From: cfe-dev <cfe-dev-bounces at lists.llvm.org <mailto:cfe-dev-bounces at lists.llvm.org>> on behalf of Owen Voorhees via cfe-dev <cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org>>
>> > Sent: Monday, May 18, 2020 10:15 AM
>> > To: cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org> <cfe-dev at lists.llvm.org <mailto:cfe-dev at lists.llvm.org>>
>> > Subject: [cfe-dev] [RFC] Upstreaming proposed change to .dia format from Swift
>> >
>> > Hi all,
>> >
>> > I was hoping to get some feedback on a proposed change to clang's serialized diagnostics format.
>> >
>> > To give some background, Swift (swift.org <http://swift.org/>) also uses this format for its diagnostics, and we recently added a new feature called "educational notes" which we'd like to start serializing. An educational note is attached to a diagnostic and is essentially just a path to associated documentation somewhere in a compiler toolchain. The documentation can then be displayed alongside emitted diagnostics to teach users about relevant language concepts or describe common problems and solutions. More details can be found at https://github.com/apple/swift/blob/master/docs/Diagnostics.md#educational-notes <https://github.com/apple/swift/blob/master/docs/Diagnostics.md#educational-notes> and an example is available at https://github.com/apple/swift/blob/master/userdocs/diagnostics/temporary-pointers.md <https://github.com/apple/swift/blob/master/userdocs/diagnostics/temporary-pointers.md>.
>> >
>> > In order to support this use case, I'd like to propose adding new records to the .dia format for educational note paths, or perhaps more generally, documentation paths. The new records would be emitted alongside the main diagnostic record, similar to the handling of ranges and fix-its. I've submitted a patch with the proposed changes here: https://reviews.llvm.org/D80126 <https://reviews.llvm.org/D80126> .
>> >
>> > The main reason we'd like to upstream this change to Clang instead of just implementing it in Swift is to avoid diverging from the established format. However, it isn't language specific and in the long term I think this is a feature that could benefit Clang's diagnostics quite a bit as well.
>> >
>> > My questions are:
>> >
>> > Is the clang community open to making a change like this that primarily benefits a downstream project (swift) in the short term?
>> > Is this a feature clang might be interested in adopting at some point?
>> >
>> >
>> > Any and all feedback is welcome! Sorry if this is the wrong place to discuss this sort of thing — it's my first time posting to the LLVM lists.
>> >
>> > Thanks,
>> > Owen
> 
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20200520/30c95a5f/attachment.html>