The docs for "PRAGMA application_id" read:
> The application_id PRAGMA is used to query or set the 32-bit unsigned
> big-endian "Application ID" integer located at [...].
However, it appears that the argument to this pragma is interpreted as a
_signed_ integer, not an unsigned integer. In particular, values outside
the ranged of a 32-bit signed integer are treated as zero, and negative
values within the range of a 32-bit signed integer are accepted but
interpreted as unsigned by file(1).
Meanwhile, magic.txt says that application_id _is_ a signed integer, and
pragma.c seems to imply it (it says that schema_version and user_version
are signed, and application_id goes through the same code path as far as
I can tell).
Is this an error in the documentation?
The following script reproduces the behavior:
#!/bin/sh
set -eu
printf '$ %s\n' 'sqlite3 --version'
sqlite3 --version
tmpdir="$(mktemp -d)"
tmpfile="${tmpdir}/test.db"
run() {
printf '\n'
rm -f "${tmpfile}"
printf 'application_id: %s\n' "$1"
printf 'SQLite says: '
printf '
PRAGMA application_id = 777;
PRAGMA application_id = %s;
PRAGMA application_id;
' "$1" | sqlite3 "${tmpfile}"
printf 'file(1) says: '
file -b "${tmpfile}"
printf 'raw bytes are: '
xxd -i -s 68 -l 4 <"${tmpfile}"
}
# Values between 0 and 2^31 - 1, inclusive, work fine.
run 0
run 3
run 2147483647
# Values larger than 2^31 - 1 are treated as zero.
run 2147483648
run 2147483649
run 4294967295
run 4294967296
run 4294967297
# Values between -2^31 and -1, inclusive, are treated inconsistently
# between SQLite and file(1).
run -2147483648
run -2147483647
run -3
# Values smaller than -2^31 are treated as zero.
run -2147483649
run -4294967295
run -4294967296
rm "${tmpfile}"
rmdir "${tmpdir}"
(End of script.)
A copy of this script and its output on my machine are available at:
https://gist.github.com/wchargin/ac2a339f002a9604018bbc3b3be59ffb
I have the following version of sqlite3...
3.11.0 2016-02-15 17:29:24 3d862f207e3adc00f78066799ac5a8c282430a5f
...running on GNU/Linux (Linux Mint 18.2).
I have searched the issue tracker and found no tickets matching
"application_id". I see one mailing list thread from 2014-02-06 that
mentions this in the context of another question, with no replies.
Best,
WC
_______________________________________________
sqlite-users mailing list
[email protected]
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
