After updating to Ubuntu 21.10 which has gcc 11 as its default compiler, sqlite builds show a compiler warning:

I just tried with sqlite-snapshot-202110132029:

In file included from /usr/include/string.h:519,
                 from sqlite3.c:13625:
In function ‘memcpy’,
    inlined from ‘sqlite3Fts5IndexQuery’ at sqlite3.c:227799:18:
/usr/include/x86_64-linux-gnu/bits/string_fortified.h:29:10: warning: ‘__builtin_memcpy’ specified bound 18446744073709551615 exceeds maximum object size 9223372036854775807 [-Wstringop-overflow=]
   29 |   return __builtin___memcpy_chk (__dest, __src, __len,
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   30 |                                  __glibc_objsize0 (__dest));
      |                                  ~~~~~~~~~~~~~~~~~~~~~~~~~~

As best I can tell, the compiler is making a guess about the largest possible buffer and the largest possible size passed in to the memcpy() call, and complaining about a theoretical problem. Assuming that the surrounding code is allocating the buffer correctly (which it looks like it should be, but I didn't really check), I don't think this is a real problem.

If I constrain nTokens to be less than UINT32_MAX (nToken is an int, sqlite3Fts5BufferSize takes a u32 size parameter), I get a slightly different warning:

In file included from /usr/include/string.h:519,
                 from sqlite3.c:13625:
In function ‘memcpy’,
    inlined from ‘sqlite3Fts5IndexQuery’ at sqlite3.c:227798:18:
/usr/include/x86_64-linux-gnu/bits/string_fortified.h:29:10: warning: ‘__builtin_memcpy’ writing 1 or more bytes into a region of size 0 overflows the destination [-Wstringop-overflow=]
   29 |   return __builtin___memcpy_chk (__dest, __src, __len,
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   30 |                                  __glibc_objsize0 (__dest));
      |                                  ~~~~~~~~~~~~~~~~~~~~~~~~~~

Checking that the destination is not NULL first seems to make that warning disappear.

This patch shows what I needed to do to make the compiler stop emitting warnings, but I don't think this is suitable for inclusion as-is. I don't feel that I have a strong enough understanding of the surrounding code to create the actual correct patch. I strongly suspect the extra conditions I added are impossible to trigger.

--- sqlite-snapshot-202110132029/sqlite3.c	2021-10-14 11:46:05.000000000 -0400
+++ sqlite-snapshot-202110132029.new/sqlite3.c	2021-10-28 11:25:49.637791053 -0400
@@ -227793,10 +227793,10 @@ static int sqlite3Fts5IndexQuery(
   /* If the QUERY_SCAN flag is set, all other flags must be clear. */
   assert( (flags & FTS5INDEX_QUERY_SCAN)==0 || flags==FTS5INDEX_QUERY_SCAN );
 
-  if( sqlite3Fts5BufferSize(&p->rc, &buf, nToken+1)==0 ){
+  if( nToken < UINT32_MAX && sqlite3Fts5BufferSize(&p->rc, &buf, nToken+1)==0 ){
     int iIdx = 0;                 /* Index to search */
     int iPrefixIdx = 0;           /* +1 prefix index */
-    if( nToken ) memcpy(&buf.p[1], pToken, nToken);
+    if( nToken && buf.p != NULL ) memcpy(&buf.p[1], pToken, nToken);
 
     /* Figure out which index to search and set iIdx accordingly. If this
     ** is a prefix query for which there is no prefix index, set iIdx to