Re: [Emacs-diffs] scratch/gnus-decoded f041412 1/7: Decode group names in newsrc files as raw-text, not utf-8

Re: [Emacs-diffs] scratch/gnus-decoded f041412 1/7: Decode group names in newsrc files as raw-text, not utf-8

[Stefan Monnier]

>  		    (decode-coding-string
> -		     (symbol-name group) 'utf-8)))
> +		     (symbol-name group) 'raw-text)))

Actually, I was misremembering: raw-text is basically like `binary`, so
the above `decoding` will only return ASCII and "eight-bit" byte-chars.
Is that the intention?


        Stefan

Re: [Emacs-diffs] scratch/gnus-decoded f041412 1/7: Decode group names in newsrc files as raw-text, not utf-8

[Eric Abrahamsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>>  		    (decode-coding-string
>> -		     (symbol-name group) 'utf-8)))
>> +		     (symbol-name group) 'raw-text)))
>
> Actually, I was misremembering: raw-text is basically like `binary`, so
> the above `decoding` will only return ASCII and "eight-bit" byte-chars.
> Is that the intention?

To be perfectly honest, I don't know what the intention is, I'm just
trying to preserve original behavior.

Consider the group named nnml:テスト.

------

In Emacs 26, this group name is the symbol
'nnml:\343\203\206\343\202\271\343\203\210, but kept in
gnus-newsrc-alist as the string version of that.

It is written to the newsrc (not newsrc.eld) file with:

(with-current-buffer buf
  (mm-disable-multibyte)
  (insert "nnml:\343\203\206\343\202\271\343\203\210")
  (let ((coding-system-for-write 'raw-text))
    (save-buffer)))
    
And read with:

(set-buffer (nnheader-find-file-noselect newsrc-file))
(setq symbol (read (current-buffer)))
(setq group (symbol-name symbol))

-----

In master, the group name is always the string
"nnml:\343\203\206\343\202\271\343\203\210".

It's written and read the same way.

-----

In scratch/gnus-decoded, the group name is the string "nnml:テスト". It
is written to newsrc with:

(with-current-buffer buf
  (insert "nnml:テスト")
  (let ((coding-system-for-write 'raw-text))
    (save-buffer)))

And read with:

(set-buffer (nnheader-find-file-noselect newsrc-file))
(setq group (read (current-buffer))
      group (decode-coding-string
	     (symbol-name group) 'raw-text))

-----

What I'm hoping is that the bytes written to the newsrc file will be
exactly the same over these three versions. This is for interoperability
with other newsreader programs.

The group name itself should go from encoded-string symbol in Emacs 26,
to encoded string in master, to decoded string in scratch/gnus-decoded.

If that's not what's happening, I would like to know!

Eric

Re: [Emacs-diffs] scratch/gnus-decoded f041412 1/7: Decode group names in newsrc files as raw-text, not utf-8

[Andreas Schwab]

On Jun 21 2019, Eric Abrahamsen <eric@ericabrahamsen.net> wrote:

> In scratch/gnus-decoded, the group name is the string "nnml:テスト". It
> is written to newsrc with:
>
> (with-current-buffer buf
>   (insert "nnml:テスト")
>   (let ((coding-system-for-write 'raw-text))
>     (save-buffer)))
>
> And read with:
>
> (set-buffer (nnheader-find-file-noselect newsrc-file))
> (setq group (read (current-buffer))
>       group (decode-coding-string
> 	     (symbol-name group) 'raw-text))

The coding rules for symbols and strings are quite different.  If you
want read a symbol, you need to write a symbol.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: [Emacs-diffs] scratch/gnus-decoded f041412 1/7: Decode group names in newsrc files as raw-text, not utf-8

[Eric Abrahamsen]

Andreas Schwab <schwab@linux-m68k.org> writes:

> On Jun 21 2019, Eric Abrahamsen <eric@ericabrahamsen.net> wrote:
>
>> In scratch/gnus-decoded, the group name is the string "nnml:テスト". It
>> is written to newsrc with:
>>
>> (with-current-buffer buf
>>   (insert "nnml:テスト")
>>   (let ((coding-system-for-write 'raw-text))
>>     (save-buffer)))
>>
>> And read with:
>>
>> (set-buffer (nnheader-find-file-noselect newsrc-file))
>> (setq group (read (current-buffer))
>>       group (decode-coding-string
>> 	     (symbol-name group) 'raw-text))
>
> The coding rules for symbols and strings are quite different.  If you
> want read a symbol, you need to write a symbol.

Actually what I want is to read/write strings. The calls to `read' made
sense when group names were always symbols, but there isn't much point
to that now. I was leaving them on general principle -- smallest viable
change, etc -- but maybe I should just bite the bullet and switch to
buffer substrings.

Eric