bug#40407: [PATCH] slow ENCODE_FILE and DECODE_FILE

[Eli Zaretskii <eliz@gnu.org>]

> From: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> Cc: Mattias Engdegård <mattiase@acm.org>,
>         40407@debbugs.gnu.org
> Date: Mon, 06 Apr 2020 19:10:48 +0900
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> -  if (BUFFERP (dst_object))
> >> +  if (EQ (dst_object, Qt))
> >> +    {
> >> +      /* Fast path for ASCII-only input and an ASCII-compatible coding:
> >> +         act as identity.  */
> >> +      Lisp_Object attrs = CODING_ID_ATTRS (coding.id);
> >> +      if (! NILP (CODING_ATTR_ASCII_COMPAT (attrs))
> >> +          && (STRING_MULTIBYTE (string)
> >> +              ? (chars == bytes) : string_ascii_p (string)))
> >> +        return string;
> 
> While using the latest master branch, I noticed this became the cause of
> decoding error.
> 
> The simple reproducible test is,
> 
> 	(decode-coding-string "&abc" 'utf-7-imap)
>         => "&abc"
> 
> like the above result, decoding utf-7-imap didn't work.
> 
> Because (coding-system-get 'utf-7-imap :ascii-compatible-p) => t.

Thanks.

> I'm not sure, 'utf-7* should be fixed as non ascii-compatible, or
> string_ascii_p() should check more strictly.

The former, since UTF-7 is definitely *not* ASCII-compatible.  Does
the patch below produce good results?

Kenichi, why was coding-type of UTF-7 systems set to 'utf-8'?
Wouldn't it be better to set it to 'utf-16'?  Or is there some
subtlety here that we should be aware of?  Do you have any comments on
the patch below?

Thanks.

diff --git a/src/coding.c b/src/coding.c
index 97a6eb9..71ff93c 100644
--- a/src/coding.c
+++ b/src/coding.c
@@ -11301,7 +11301,10 @@ DEFUN ("define-coding-system-internal", Fdefine_coding_system_internal,
 	  CHECK_CODING_SYSTEM (val);
 	}
       ASET (attrs, coding_attr_utf_bom, bom);
-      if (NILP (bom))
+      if (NILP (bom)
+	  /* UTF-7 has :coding-type set to 'utf-8' (why not
+	     'utf-16'?), but it is definitely NOT ASCII-compatible.  */
+	  && !EQ (name, Qutf_7) && !EQ (name, Qutf_7_imap))
 	ASET (attrs, coding_attr_ascii_compat, Qt);
 
       category = (CONSP (bom) ? coding_category_utf_8_auto
@@ -11673,6 +11676,9 @@ syms_of_coding (void)
   DEFSYM (Qutf_8_unix, "utf-8-unix");
   DEFSYM (Qutf_8_emacs, "utf-8-emacs");
 
+  DEFSYM (Qutf_7, "utf-7");
+  DEFSYM (Qutf_7_imap, "utf-7-imap");
+
 #if defined (WINDOWSNT) || defined (CYGWIN)
   /* No, not utf-16-le: that one has a BOM.  */
   DEFSYM (Qutf_16le, "utf-16le");