Re: [PATCH for Dlang support] d: change the return value of yylex() from int to TokenKind

[edi33416]

> On Sep 28, 2020, at 10:13 PM, Adela Vais <adela.vais99@gmail.com> wrote:
> 
> Hello Akim,
> 
> În lun., 28 sept. 2020 la 20:34, Akim Demaille <akim@lrde.epita.fr> a scris:
> 
>> 
>> You mean bison.texi?  That's great!  The README file about the
>> implementation of the D skeleton is scheduled to be deleted.
>> 
>> 
> Yes, bison.texi.
> 
> 
>> 
>>>> There's something fishy happening on the CI: the examples are failing.
>>>> It is not a recent change in the skeleton, it is a change in the
>> toolchain.
>>>> We now get errors from address sanitizer.  Have a look at
>>>> https://travis-ci.org/github/akimd/bison/jobs/729867686 for instance.
>>>> 
>>> 
>>> Is there something I can do about this error, or it's outside of my
>> reach?
>> 
>> I have no idea.  You should try to reproduce the recipes from .travis.yml.
>> 
>>> I tried to replicate it, but I get nothing when I run
>>> ldc2 -fsanitize=address calc.d
>>> I found online only ldc examples for -fsanitize=address for compiling
>> Dlang.
>>> Can you please give me the command the tests use (if they are using
>>> something else)?
>> 
>> I could not reproduce this either on this machine (a macOS).  Actually,
>> since travis is using dmd, not gcc for d (but maybe we should try that
>> too), I don't even understand how it's possible that ASAN finds things.
>> 
> 
> I'll look into this, I really don't know what happens there.
> 
> 
>> ldc2 is mature enough?  Maybe the CI should try not online dmd, but also
>> GCC and LLVM's support for D.
>> 
> 
> I don't know, but I'll ask the D community and give an update on this.

Hello,

Yes, ldc2 and gdc are mature enough.
ldc2 and gdc are the compilers recommended for production use.

I think it might be a good idea to update travis to use ldc2.

Edi