8:56 p.m.
This is a bug in the built-in function 'forward-comment' of later XEmacs
versions. I already reported this bug 10 months ago:
- Bug in `forward-comment' (only 21.4, 21.1 is ok):
Test case:
- load mode that has comment-start = "--" (ada-mode, vhdl-mode)
- (forward-comment '-1) after a line that starts with an odd
number of "-", e.g.
--- this is a legal comment
^ cursor will be positioned here
This bug is severe because it breaks indentation as well!
Reto
michael methfessel wrote:
Hi,
thank you for writing vhdl-mode, a very usefool tool. However, there
is one annoying feature which is presumably a bug. When I use three
hyphens to start a comment (this happens to be my personal convention)
the indentation is faulty in the following lines. Here is my vhdl file:
library IEEE;
use IEEE.Std_Logic_1164.all;
entity mips4kep is
port (cmd: in integer;
eb_avalid: out std_logic;
eb_write: out std_logic;
eb_a: out std_logic_vector (29 downto 0);
eb_rdata: in std_logic_vector (31 downto 0);
eb_wdata: out std_logic_vector (31 downto 0)
);
end;
-- fmi --
architecture fmi of mips4kep is
attribute Foreign of fmi:architecture is
"libfmi:mips4kep";
begin
end fmi;
In this form, indentation works. But if I write the comment as
--- fmi --
things go wrong starting with line "architecture fmi of mips4kep is".
To kreis this ein I switched on vhdl-echo-syntactic-information-p:
-- fmi --
syntax: ((comment) (statement . 47)), indent= -1000
architecture fmi of mips4kep is
syntax: ((statement . 47)), indent= 0
--- fmi --
syntax: ((comment) (statement . 47)), indent= -1000
architecture fmi of mips4kep is
syntax: ((statement-cont . 347)), indent= 4
So, "architecture.." is indented by 4 if the comment starts with
"---".
I tried looking into the lisp, but I don't think I would be able
to find the reason myself.
Thanks for any help, Michael
10:15 a.m.
>>>> "Reto" == Reto Zimmermann
<reto(a)synopsys.com> writes:
Reto> This is a bug in the built-in function 'forward-comment' of
Reto> later XEmacs versions. I already reported this bug 10
Reto> months ago:
Do you have a fix? I've put in several hours on this and don't have
one. Sorry.
--
Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
My nostalgia for Icon makes me forget about any of the bad things. I don't
have much nostalgia for Perl, so its faults I remember. Scott Gilbert c.l.py
6:30 p.m.
Stephen J. Turnbull wrote:
Reto> This is a bug in the built-in function
'forward-comment' of
Reto> later XEmacs versions. I already reported this bug 10
Reto> months ago:
Do you have a fix? I've put in several hours on this and don't have
one. Sorry.
No, I don't have a fix, because this is a built-in function and I've
never looked into nor written any source code for XEmacs nor GNU Emacs.
I'm a lisp-level programmer only.
But this function worked well before (e.g. in 21.1) so it must be a
recent change that broke it.
Reto
4:16 a.m.
>>>> "Reto" == Reto Zimmermann
<reto(a)synopsys.com> writes:
Reto> But this function worked well before (e.g. in 21.1) so it
Unfortunately (forward-comment -1) uses a less-than-transparent
"intelligent best effort" algorithm. We've already tried reverting
syntax.c to 21.1 (for other reasons), and borked stuff appeared all
over the place. That's what the traffic about broken 21.4.9 Windows
[1,2] was about (I never released a mainline XEmacs with that code in
it, but Andy did release a couple of Windows versions).
Reto> must be a recent change that broke it.
21.2 (the development branch preceding 21.4) forked from 21.1 in late
1997, and there have been two major overhauls of the syntax code in
21.2 before the release of 21.4, plus a large number of changes since.
Even by Emacs standards, that's not "recent". :-)
bash-2.05b$ diff -u {../21.1/,}src/syntax.c | wc
2122 9512 69165
bash-2.05b$ wc src/syntax.c
2332 8696 68132 src/syntax.c
Unless somebody is willing to volunteer to do some intensive work on
that code, it's going to take time.
One straightforward experiment would be to rip out the current
"beautiful mind" code for negative arguments to forward-comment, and
replace it with a single parse from the beginning of the file. The
correlation between languages whose programmers tend to produce large
source files (> 100kB) and languages which favor single-top-level-defun
source files is high enough that it's probably on average cheaper to
do it that way. :-( However, it could cause a noticable performance
hit for large Lisp and C files when doing (forward-comment -1) near
the end of the file.
This will probably not be forthcoming from me for at least two weeks,
and possibly more like beginning of December, though.
--
Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
My nostalgia for Icon makes me forget about any of the bad things. I don't
have much nostalgia for Perl, so its faults I remember. Scott Gilbert c.l.py
8067
days inactive
8071
days old
3 comments
2 participants
participants (2)
-
Reto Zimmermann -
Stephen J. Turnbull