bug#5033: 23.1; antlr-mode: antlr-indent-command broken for long grammars

bug#5033: 23.1; antlr-mode: antlr-indent-command broken for long grammars

[Mr Kludge]

[-- Attachment #1.1: Type: text/plain, Size: 1054 bytes --]

Hello,

I am running into an antlr-mode bug when indenting long grammars.
Specifically, when the point in the grammar is higher than
antlr-slow-cache-diff-threshold indenting a grammar rule fails in
calling c-where-wrt-brace-construct through beginning-of-defun.

To trigger:
Write a longinsh grammar (more than 5k chars with the default setting)
or simply set antlr-slow-cache-diff-threshold to 0 and then try to
indent in a grammar rule.
See attached silly.g for an example.

Indentation breaks, with the following error:
c-where-wrt-brace-construct: c-beginning-of-decl-1 returned label

Backtrace:
  c-where-wrt-brace-construct()
  c-beginning-of-defun(1)
  funcall(c-beginning-of-defun 1)
  beginning-of-defun-raw(nil)
  ad-Orig-beginning-of-defun(nil)
  beginning-of-defun()
  antlr-syntactic-context()
  antlr-indent-line()
  antlr-indent-command(nil)
  call-interactively(antlr-indent-command nil nil)

Solution:
A simple solution (kludge) is to modify antlr-syntactic-context to not call
beginning-of-defun in antlr-mode.el.
See attached patch.

[-- Attachment #1.2: Type: text/html, Size: 1170 bytes --]

[-- Attachment #2: antlr-mode.el.patch --]
[-- Type: application/octet-stream, Size: 616 bytes --]

--- lisp/progmodes/antlr-mode.el.old	2009-11-24 09:50:22.000000000 -0500
+++ lisp/progmodes/antlr-mode.el	2009-11-24 09:51:24.000000000 -0500
@@ -1111,9 +1111,7 @@
 	(setq state (parse-partial-sexp (cadr antlr-slow-context-cache) orig
 					nil nil
 					(cddr antlr-slow-context-cache)))
-      (if (>= orig antlr-slow-cache-diff-threshold)
-	  (beginning-of-defun)
-	(goto-char (point-min)))
+    (goto-char (point-min))
 ;;;      (cond ((and diff (< diff 0)) (incf antlr-statistics-full-neg))
 ;;;	    ((and diff (>= diff 3000)) (incf antlr-statistics-full-diff))
 ;;;	    (t (incf antlr-statistics-full-other)))

[-- Attachment #3: silly.g --]
[-- Type: application/octet-stream, Size: 294 bytes --]

// antlr-mode indentation bug
// to trigger:
// (setq antlr-slow-cache-diff-threshold 0)
// indent anywhere in a rule
// errors [emacs 23.1] with
//  c-where-wrt-brace-construct: c-where-wrt-brace-construct: c-beginning-of-decl-1 returned label
grammar silly;
foo
    : (bar)
    | (baz)
    ;

bug#5033: 23.1; antlr-mode: antlr-indent-command broken for long grammars

[Chong Yidong]

> I am running into an antlr-mode bug when indenting long grammars.
> Specifically, when the point in the grammar is higher than
> antlr-slow-cache-diff-threshold indenting a grammar rule fails in
> calling c-where-wrt-brace-construct through beginning-of-defun.

> Solution:
> A simple solution (kludge) is to modify antlr-syntactic-context to not
> call beginning-of-defun in antlr-mode.el.  See attached patch.

I am not able to test this solution, because I don't work with antlr
files (and I doubt any of the other core Emacs maintainers do either).
I'm CC'ing the author of antlr-mode, Christoph Wedler, though we haven't
heard from him in a while.

In the meantime, could you try to explain what removing the
beginning-of-defun call in antlr-syntactic-context might break?  (Surely
that call is there for a reason.)

bug#5033: 23.1; antlr-mode: antlr-indent-command broken for long grammars

[Andrew Hyatt]

I cannot reproduce this on Emacs 25.  I loaded the provided silly.g, set
the antrl-slow-cache-diff-threshold to 0, and indented each line without
any issues, at many different points on the line.

Someone can let me know if they can still reproduce this, otherwise I'll
close this out.

Chong Yidong <cyd@stupidchicken.com> writes:

>> I am running into an antlr-mode bug when indenting long grammars.
>> Specifically, when the point in the grammar is higher than
>> antlr-slow-cache-diff-threshold indenting a grammar rule fails in
>> calling c-where-wrt-brace-construct through beginning-of-defun.
>
>> Solution:
>> A simple solution (kludge) is to modify antlr-syntactic-context to not
>> call beginning-of-defun in antlr-mode.el.  See attached patch.
>
> I am not able to test this solution, because I don't work with antlr
> files (and I doubt any of the other core Emacs maintainers do either).
> I'm CC'ing the author of antlr-mode, Christoph Wedler, though we haven't
> heard from him in a while.
>
> In the meantime, could you try to explain what removing the
> beginning-of-defun call in antlr-syntactic-context might break?  (Surely
> that call is there for a reason.)

bug#5033: 23.1; antlr-mode: antlr-indent-command broken for long grammars

[Stefan Kangas]

Andrew Hyatt <ahyatt@gmail.com> writes:

> I cannot reproduce this on Emacs 25.  I loaded the provided silly.g, set
> the antrl-slow-cache-diff-threshold to 0, and indented each line without
> any issues, at many different points on the line.
>
> Someone can let me know if they can still reproduce this, otherwise I'll
> close this out.

It was never closed at the time, 3 years ago, so I'm closing this now.

If anyone can still reproduce this, please reopen the bug report.

Best regards,
Stefan Kangas