bug#66845: 29.1; cmake-ts-mode indentation broken

[Randy Taylor]

On Tuesday, November 7th, 2023 at 21:52, Yuan Fu <casouri@gmail.com> wrote:
> 
> > On Nov 7, 2023, at 9:21 AM, Randy Taylor dev@rjt.dev wrote:
> > 
> > On Tuesday, November 7th, 2023 at 03:25, Yuan Fu casouri@gmail.com wrote:
> > 
> > > IMO it makes more sense to use prev-sibling:
> > > 
> > > ,@(ignore-errors
> > > (treesit-query-capture 'cmake '((body) @capture))
> > > `(((parent-is "body") prev-sibling 0)))
> > > 
> > > This would allow users to manually adjust the indentation of a line and 
> > > have the rest of the body follow that.
> > 
> > This would be much more elegant but unfortunately that and my previous 
> > patch failed to account for the following:
> > 
> > 1 if(TRUE) # Comment.
> > 2 endif()
> > 
> > And place POINT at the end of line 1 and hit RET. It will indent to the 
> > start of the comment :(. However, you gave me an idea: use the grand-parent 
> > as the anchor (which in this case is the if - exactly what we want). I've 
> > attached a new patch which accounts for that.
> 
> 
> Ah, I was a bit rusty too. The approach I mentioned should also have a rule 
> for the first sibling which anchors on the parent, while the rest siblings 
> anchor on the previous sibling. But anchoring all on the parent is equally 
> valid.

I see.

Could you (or someone else) install the patch to emacs-29?

> 
> > Juan, thanks for testing the first patch. Please give this new one a whirl 
> > if you're able to.
> > 
> > > As for why parent-bol returns the BOL of the comment line, that’s 
> > > expected: The parent is body, and body starts at the comment, so of 
> > > course the beginning of the parent line is the beginning of the comment 
> > > line. In the image below, the highlighted portion marks the body node.
> > > 
> > > Some tree-sitter grammar would mark the beginning of body at the end of 
> > > if(TRUE), but what tree-sitter-cmake does here is equally valid.
> > 
> > Thanks, I understand what's happening now. Whenever line_comment is NOT the 
> > first child, the beginning of body is at the end of if(TRUE). If 
> > line_comment is the first child, then the beginning of body is at the start 
> > of the comment. I was expecting them to behave the same - that's why I was 
> > so confused, and I don't really understand why it was done that 
> > way.<0001-Fix-cmake-ts-mode-indentation-Bug-66845.patch>
> 
> 
> You’re right. That’s a strange behavior. We should probably report to 
> tree-sitter-cmake.

I will try to get around to that sometime next week.