Re: [PATCH 0/4] Fix caret errors

[Akim Demaille]

> Le 22 avr. 2019 à 10:19, Hans Åberg <address@hidden> a écrit :
> 
>> On 22 Apr 2019, at 07:29, Akim Demaille <address@hidden> wrote:
>> 
>>> The only safe way is to replace the visible characters in the quoted string 
>>> with spaces, and then use an initial portion of that. Then you can count 
>>> characters as you like, probably tabs as 1.
>> 
>> What do you mean by "the only safe way"?  Paul's proposal a la diff -T,
>> or GCC9's approach both seem to work properly.
> 
> In case somebody mixes tabs and spaces, like four space for tabs and mixed 
> with spaces?

Ah, you mean when tabs are not 8 spaces.  Yes, you are right, in
that case, it won't work.

The GNU coding standards make it clear that tabs are 8 spaces.
(https://www.gnu.org/prep/standards/standards.html#index-formatting-error-messages),
 and so far no-one complained about the handling of tabs
in Bison (except me :-).  Eventually, if there are complains, and if
GCC9 also makes the choice to be independent of tab width by smashing
them to spaces, I'll adjust Bison.


>>> Does not work with Unicode, though. In your example
>>>> input.y:15.4-17: warning: empty rule without %empty [-Wempty-rule]
>>>> e: {∇⃗×𝐸⃗ = -∂𝐵⃗/∂t}
>>>>  ^~~~~~~~~~~~~~
>>> here, the U+20D7 COMBINING RIGHT ARROW ABOVE for the nabla combines, but 
>>> others do not.
>> 
>> What do you mean?  I see it perfectly: arrows are combined to E and B.
>> It might be something on your mail reader side?  
> 
> Right, it depends on the renderer, so if you want a different line with 
> matching spacing, that is not really possible.

Agreed.  I assume that people use a fixed width font, or disable
caret diagnostics.


>> For some reason, in
>> Apple's terminal, both nabla and B are correctly "accented", but not
>> the E.  All are correct in the mailer though.
> 
> I use Mail on MacOS 10.13.

Well, then they fixed that in 10.14(.4).