bug#47468: 27.1; cc-mode: Got incorrect indentaton for C++ lambda function.

bug#47468: 27.1; cc-mode: Got incorrect indentaton for C++ lambda function.

[Jianbin Kang]

[-- Attachment #1: Type: text/plain, Size: 1015 bytes --]

The c++ file to reproduce it:
```c++
#include <stdio.h>
#include <functional>

struct PP {
    const char *        name;
};

static void print(const char *name, std::function<void(const char *)> f)
{
    f(name);
}

int main(int argc, char *argv[])
{
    PP A = { "Jim" };
    PP *a = &A;

    print(A.name, [](const char *name) {
        printf("%s\n", name);           // Syntactic ((inlambda) ...
    });
    print(a->name, [=](const char *name) {
            printf("%s\n", name);       // Syntactic((arglist-cont-nonempty
...
        });

    return 0;
}

```

If I put cursor in lambda body of first print call, and run
'c-show-syntactic-information', it shows 'Syntactic analysis: ((inlambda)
(defun-block-intro 258))', which is correct and I get good indentation.

But in second print call, I get 'Syntactic analysis:
((arglist-cont-nonempty 337 342) (statement-block-intro 337))' and the
indentation is bad.

My emacs version is 27.1. I can reproduce the problem in both Linux and
Windows with 'emacs -Q'.

[-- Attachment #2: Type: text/html, Size: 1320 bytes --]

bug#47468: 27.1; cc-mode: Got incorrect indentaton for C++ lambda function.

[Lars Ingebrigtsen]

[-- Attachment #1: Type: text/plain, Size: 1066 bytes --]

Jianbin Kang <kjbmail@gmail.com> writes:

> int main(int argc, char *argv[])
> {
>     PP A = { "Jim" };
>     PP *a = &A;
>
>     print(A.name, [](const char *name) {
>         printf("%s\n", name);           // Syntactic ((inlambda) ... 
>     });
>     print(a->name, [=](const char *name) {
>             printf("%s\n", name);       // Syntactic((arglist-cont-nonempty ...
>         });
>
>     return 0;

[...]

> If I put cursor in lambda body of first print call, and run 'c-show-syntactic-information',
> it shows 'Syntactic analysis: ((inlambda) (defun-block-intro 258))', which is correct and
> I get good indentation.
>
> But in second print call, I get 'Syntactic analysis:
> ((arglist-cont-nonempty 337 342) (statement-block-intro 337))' and the
> indentation is bad.

I can reproduce this in Emacs 28, too.  Poking around a bit, it seems
like it's getting tripped up by the "a->name" bit?  Changing that to
a.name (for instance) makes it recognise the second form as a lambda
function...

However, the first form is indented wrong for me in Emacs 28:


[-- Attachment #2: Type: image/png, Size: 25838 bytes --]

[-- Attachment #3: Type: text/plain, Size: 245 bytes --]


This is indented correctly in Emacs 27.1, so it seems like this has
regressed between Emacs 27.1 and Emacs 28.

I've added Alan to the CCs.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#47468: 27.1; cc-mode: Got incorrect indentaton for C++ lambda function.

[Alan Mackenzie]

Hello, Jianbin.

Thank you indeed for taking the trouble to report this bug, and thank
you even more for cutting the test case down to a concise, easy to work
with file.

On Mon, Mar 29, 2021 at 20:47:52 +0800, Jianbin Kang wrote:
> The c++ file to reproduce it:
> ```c++
> #include <stdio.h>
> #include <functional>

> struct PP {
>     const char *        name;
> };

> static void print(const char *name, std::function<void(const char *)> f)
> {
>     f(name);
> }

> int main(int argc, char *argv[])
> {
>     PP A = { "Jim" };
>     PP *a = &A;

>     print(A.name, [](const char *name) {
>         printf("%s\n", name);           // Syntactic ((inlambda) ...
>     });
>     print(a->name, [=](const char *name) {
>             printf("%s\n", name);       // Syntactic((arglist-cont-nonempty
> ...
>         });

>     return 0;
> }

> ```

> If I put cursor in lambda body of first print call, and run
> 'c-show-syntactic-information', it shows 'Syntactic analysis: ((inlambda)
> (defun-block-intro 258))', which is correct and I get good indentation.

> But in second print call, I get 'Syntactic analysis:
> ((arglist-cont-nonempty 337 342) (statement-block-intro 337))' and the
> indentation is bad.

Yes.  What is happening is that CC Mode is falsely recognising the -> in
a->name as the trailing return type of the lambda expression.

I think the following patch should fix it.  Could I ask you, please, to
apply this patch to ..../emacs/lisp/progmodes/cc-engine.el, byte-compile
the file, and try it out on your real C++ source code.  In the unlikely
event you want help with the patching or byte-compilation, feel free to
send me private email.  After this, please either confirm to us that the
problem is fixed, or tell us what is still not working.  Thanks!



diff -r 92a4592886a1 cc-engine.el
--- a/cc-engine.el	Sun Apr 25 17:26:38 2021 +0000
+++ b/cc-engine.el	Sat May 15 14:51:56 2021 +0000
@@ -12269,7 +12269,7 @@
 		       (save-excursion
 			 (while
 			     (progn
-			       (c-syntactic-skip-backward "^;=}>" closest-lim t)
+			       (c-syntactic-skip-backward "^;=,}>" closest-lim t)
 			       (and (eq (char-before) ?>)
 				    (c-backward-token-2)
 				    (not (looking-at c-haskell-op-re)))))


> My emacs version is 27.1. I can reproduce the problem in both Linux and
> Windows with 'emacs -Q'.

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#47468: 27.1; cc-mode: Got incorrect indentaton for C++ lambda function.

[Vladimir Lichevsky]

Not the original poster, but I tried this patch on recent build of
Emacs from master branch (commit
567c31121fdef6bdc8b645999a6ca1d994378c89) and it looks like the patch
fixes the bug.

bug#47468: 27.1; cc-mode: Got incorrect indentaton for C++ lambda function.

[Alan Mackenzie]

Hello, Jianbin and Lars.

Some while has passed since the bug was opened, and since the patch I
suggested seems to fix it, I am now closing this bug.

-- 
Alan Mackenzie (Nuremberg, Germany).


On Sat, May 15, 2021 at 14:57:45 +0000, Alan Mackenzie wrote:
> Hello, Jianbin.

> Thank you indeed for taking the trouble to report this bug, and thank
> you even more for cutting the test case down to a concise, easy to work
> with file.

> On Mon, Mar 29, 2021 at 20:47:52 +0800, Jianbin Kang wrote:
> > The c++ file to reproduce it:
> > ```c++
> > #include <stdio.h>
> > #include <functional>

> > struct PP {
> >     const char *        name;
> > };

> > static void print(const char *name, std::function<void(const char *)> f)
> > {
> >     f(name);
> > }

> > int main(int argc, char *argv[])
> > {
> >     PP A = { "Jim" };
> >     PP *a = &A;

> >     print(A.name, [](const char *name) {
> >         printf("%s\n", name);           // Syntactic ((inlambda) ...
> >     });
> >     print(a->name, [=](const char *name) {
> >             printf("%s\n", name);       // Syntactic((arglist-cont-nonempty
> > ...
> >         });

> >     return 0;
> > }

> > ```

> > If I put cursor in lambda body of first print call, and run
> > 'c-show-syntactic-information', it shows 'Syntactic analysis: ((inlambda)
> > (defun-block-intro 258))', which is correct and I get good indentation.

> > But in second print call, I get 'Syntactic analysis:
> > ((arglist-cont-nonempty 337 342) (statement-block-intro 337))' and the
> > indentation is bad.

> Yes.  What is happening is that CC Mode is falsely recognising the -> in
> a->name as the trailing return type of the lambda expression.

> I think the following patch should fix it.  Could I ask you, please, to
> apply this patch to ..../emacs/lisp/progmodes/cc-engine.el, byte-compile
> the file, and try it out on your real C++ source code.  In the unlikely
> event you want help with the patching or byte-compilation, feel free to
> send me private email.  After this, please either confirm to us that the
> problem is fixed, or tell us what is still not working.  Thanks!



> diff -r 92a4592886a1 cc-engine.el
> --- a/cc-engine.el	Sun Apr 25 17:26:38 2021 +0000
> +++ b/cc-engine.el	Sat May 15 14:51:56 2021 +0000
> @@ -12269,7 +12269,7 @@
>  		       (save-excursion
>  			 (while
>  			     (progn
> -			       (c-syntactic-skip-backward "^;=}>" closest-lim t)
> +			       (c-syntactic-skip-backward "^;=,}>" closest-lim t)
>  			       (and (eq (char-before) ?>)
>  				    (c-backward-token-2)
>  				    (not (looking-at c-haskell-op-re)))))


> > My emacs version is 27.1. I can reproduce the problem in both Linux and
> > Windows with 'emacs -Q'.

> -- 
> Alan Mackenzie (Nuremberg, Germany).