bug#23397: 25.0.92; assertion failure auto-reverting a file being overwritten

[Noam Postavsky <npostavs@gmail.com>]

found 23397 26.1
tags 23397 + confirmed
quit

Ken Raeburn <raeburn@permabit.com> writes:

> There's code there doing lseek and read calls; we should be able to keep
> track of the last file position we read from, if that's not already
> derivable from existing variables visible at that point.

This seems to fix the scenario you gave in #17.  I don't quite
understand what the overlap expression is trying to compute though.

--- i/src/fileio.c
+++ w/src/fileio.c
@@ -3849,6 +3849,8 @@ because (1) it preserves some marker positions and (2) it puts less data
 	  goto handled;
 	}
 
+      off_t file_end_pos = st.st_size;
+
       /* Count how many chars at the end of the file
 	 match the text at the end of the buffer.  But, if we have
 	 already found that decoding is necessary, don't waste time.  */
@@ -3879,6 +3881,8 @@ because (1) it preserves some marker positions and (2) it puts less data
 	      total_read += nread;
 	    }
 
+          file_end_pos = max (curpos - trial + total_read, file_end_pos);
+
 	  /* Scan this bufferful from the end, comparing with
 	     the Emacs buffer.  */
 	  bufpos = total_read;
@@ -3931,7 +3935,7 @@ because (1) it preserves some marker positions and (2) it puts less data
 	  /* Don't try to reuse the same piece of text twice.  */
 	  overlap = (same_at_start - BEGV_BYTE
 		     - (same_at_end
-			+ (! NILP (end) ? end_offset : st.st_size) - ZV_BYTE));
+			+ (! NILP (end) ? end_offset : file_end_pos) - ZV_BYTE));
 	  if (overlap > 0)
 	    same_at_end += overlap;
 	  same_at_end_charpos = BYTE_TO_CHAR (same_at_end);