From: Paul Eggert <eggert@cs.ucla.edu>
Date: Sat, 12 Jul 2025 17:03:26 +0000 (-0700)
Subject: Improve inserted file coding system finding
X-Git-Url: http://git.eshelyaron.com/gitweb/?a=commitdiff_plain;h=e789d236a3fcc7c26c8c7c5376f5b7beb6fabcf7;p=emacs.git

Improve inserted file coding system finding

* src/fileio.c (Finsert_file_contents): When inserting a file into
a nonempty buffer, improve the heuristic for determining the
file’s coding system by not trusting lseek+SEEK_END, which is
unreliable in /proc or when the file is mutating.

(cherry picked from commit 6a9dbed40ccebc18fda6078058b1978e29ef98fa)
---

diff --git a/src/fileio.c b/src/fileio.c
index cbdfeacf8e4..5d383a7b44d 100644
--- a/src/fileio.c
+++ b/src/fileio.c
@@ -4289,12 +4289,38 @@ by calling `format-decode', which see.  */)
 		{
 		  off_t tailoff = emacs_fd_lseek (fd, - 3 * 1024, SEEK_END);
 		  if (tailoff < 0)
-		    seekable = false;
-		  else if (1024 < tailoff)
 		    {
-		      ptrdiff_t ntail = emacs_fd_read (fd, read_buf + 1024,
-						       3 * 1024);
-		      nread = ntail < 0 ? ntail : 1024 + ntail;
+		      seekable = false;
+		      tailoff = nread;
+		    }
+
+		  /* When appending the last 3 KiB, read until EOF
+		     without trusting tailoff, as the file may be in
+		     /proc or be mutating.  */
+		  nread = 1024;
+		  for (;;)
+		    {
+		      ptrdiff_t r = emacs_fd_read (fd, read_buf + nread,
+						   sizeof read_buf - nread);
+		      if (r <= 0)
+			{
+			  if (r < 0)
+			    nread = r;
+			  else
+			    file_size_hint = tailoff;
+			  break;
+			}
+		      tailoff += r;
+		      nread += r;
+		      bool eof = nread < sizeof read_buf;
+		      if (4 * 1024 < nread)
+			{
+			  memmove (read_buf + 1024,
+				   read_buf + nread - 3 * 1024, 3 * 1024);
+			  nread = 4 * 1024;
+			}
+		      if (eof)
+			break;
 		    }
 		}