GNU bug report logs -
#22079
CC Mode 5.33 (C++/l); C++ cache inconsistency
Previous Next
Full log
View this message in rfc822 format
Hello, Michael.
Thanks for another bug report.
On Wed, Dec 02, 2015 at 11:28:00PM -0500, Michael Welsh Duggan wrote:
> Steps to reproduce:
> Using the attached traits.hpp file:
> emacs -Q traits.hpp
> M-x c-toggle-parse-state-debug
> M->
> In *Messages* I find:
> c-parse-state inconsistency at 4541: using cache: (4451 (2110 . 4210)), from scratch: (4451 (2110 . 4210) 1868 (123 . 1830))
> Old state:
> (setq c-state-cache ’((4451 . 5298)) c-state-cache-good-pos 5298 c-state-nonlit-pos-cache ’(3001) c-state-nonlit-pos-cache-limit 3001 c-state-semi-nonlit-pos-cache ’(3001) c-state-semi-nonlit-pos-cache-limit 3001 c-state-brace-pair-desert nil c-state-point-min 1 c-state-point-min-lit-type nil c-state-point-min-lit-start nil c-state-min-scan-pos 1 c-state-old-cpp-beg (copy-marker 5325) c-state-old-cpp-end (copy-marker 5351) c-parse-state-point 5325)
> c-parse-state inconsistency at 4661: using cache: (4660 4451 (2110 . 4210)), from scratch: (4660 4451 (2110 . 4210) 1868 (123 . 1830))
> Old state:
> (setq c-state-cache ’(4451 (2110 . 4210)) c-state-cache-good-pos 4452 c-state-nonlit-pos-cache ’(3001) c-state-nonlit-pos-cache-limit 3001 c-state-semi-nonlit-pos-cache ’(3001) c-state-semi-nonlit-pos-cache-limit 3001 c-state-brace-pair-desert nil c-state-point-min 1 c-state-point-min-lit-type nil c-state-point-min-lit-start nil c-state-min-scan-pos 1 c-state-old-cpp-beg nil c-state-old-cpp-end nil c-parse-state-point 4541)
[ .... ]
I see you are using a relatively recent snapshot of the savannah master
or emacs-25 branch (from ~26th November).
The cause of the problem is not hard to track down. On 11th November, I
committed a fix "CC Mode: Respect users' settings of
open-paren-in-column-0-is-defun-start.", which does precisely what it
says. Previously, I had bound open-paren-in-column-0-is-defun-start to
nil around certain primitive calls for the sake of accurate scanning.
The problem with binding o-p-i-c-0-i-d-s to nil is that it makes CC Mode
very slow, particularly in big files, where continually scanning from
the beginning of the file is sub-optimal.
In the scenario you documented, we have ....
> namespace std { <=============== B
>
> static_assert(ipfix::traits::is_same<uint8_t, unsigned char>,
> "Assumption that uint8_t is unsigned char is false");
>
> template<>
> struct char_traits<ipfix::traits::enable_if<
> ipfix::traits::is_signed<char>, unsigned char>>
> { <=============== A
> typedef unsigned char char_type;
> typedef char_traits<char>::int_type int_type;
> typedef char_traits<char>::off_type off_type;
> typedef char_traits<char>::pos_type pos_type;
> typedef char_traits<char>::state_type state_type;
[ .... ]
The actual top level (I think) is the brace at B. CC Mode has, however,
taken the brace at A (in column 0) to be a top level brace.
This is a difficult problem to resolve satisfactorally, in general.
What you could do is set open-paren-etc. to nil, or even set it to nil
buffer locally for some files. When I tried this on traits.hpp, it
scanned without any parse state inconsistencies.
> Emacs : GNU Emacs 25.1.50.1 (x86_64-pc-linux-gnu, X toolkit)
> of 2015-11-26
> Package: CC Mode 5.33 (C++/l)
> Buffer Style: SiLK
> c-emacs-features: (pps-extended-state col-0-paren posix-char-classes gen-string-delim gen-comment-delim syntax-properties 1-bit)
[ full CC Mode dump snipped, but appreciated. ]
> --
> Michael Welsh Duggan
> (md5i <at> md5i.com)
--
Alan Mackenzie (Nuremberg, Germany).
This bug report was last modified 9 years and 293 days ago.
Previous Next
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.