Re: Non-compliant SASLprep implementation for ASCII characters

Hello Michael,

19.03.2026 06:25, Michael Paquier wrote:
> Applied the result for the module, to have at least the coverage part.
> The last piece is refreshed, and attached for now.
> --

When running make check for src/test/modules/test_saslprep under Valgrind,
I've discovered:
# --- /pgtest/postgresql.git/src/test/modules/test_saslprep/expected/test_saslprep.out 2026-04-12 07:44:47.090517505 +0300
# +++ /pgtest/postgresql.git/src/test/modules/test_saslprep/results/test_saslprep.out 2026-04-12 08:03:29.353348951 +0300
# @@ -2,151 +2,7 @@
#  CREATE EXTENSION test_saslprep;
#  -- Incomplete UTF-8 sequence.
#  SELECT test_saslprep('\xef');
# -  test_saslprep
# ------------------
# - (,INVALID_UTF8)
# -(1 row)
# -
...
-
-DROP EXTENSION test_saslprep;
+server closed the connection unexpectedly
+   This probably means the server terminated abnormally
+   before or while processing the request.
+connection to server was lost

src/test/modules/test_saslprep/log/postmaster.log
2026-04-12 08:03:26.064 EEST postmaster[1043298] LOG:  database system is ready to accept connections
2026-04-12 08:03:26.078 EEST dead-end client backend[1043325] [unknown] FATAL:  the database system is starting up
==00:00:00:04.413 1043360== Invalid read of size 1
==00:00:00:04.413 1043360==    at 0x484F234: strlen (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==00:00:00:04.413 1043360==    by 0x5277C6F: strlcpy (strlcpy.c:24)
==00:00:00:04.413 1043360==    by 0x48648CF: test_saslprep (test_saslprep.c:87)
==00:00:00:04.413 1043360==    by 0x541A45: ExecInterpExpr (execExprInterp.c:979)
==00:00:00:04.413 1043360==    by 0x5446A1: ExecInterpExprStillValid (execExprInterp.c:2301)
==00:00:00:04.413 1043360==    by 0x5AD1FD: ExecEvalExprNoReturn (executor.h:433)
==00:00:00:04.413 1043360==    by 0x5AD2BB: ExecEvalExprNoReturnSwitchContext (executor.h:474)
==00:00:00:04.413 1043360==    by 0x5AD31C: ExecProject (executor.h:506)
==00:00:00:04.413 1043360==    by 0x5AD53F: ExecResult (nodeResult.c:135)
==00:00:00:04.413 1043360==    by 0x55F5E4: ExecProcNodeFirst (execProcnode.c:469)
==00:00:00:04.413 1043360==    by 0x55108F: ExecProcNode (executor.h:327)
==00:00:00:04.413 1043360==    by 0x554149: ExecutePlan (execMain.c:1736)
==00:00:00:04.413 1043360==  Address 0x7439295 is 6,917 bytes inside a block of size 8,192 alloc'd
==00:00:00:04.413 1043360==    at 0x4846828: malloc (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==00:00:00:04.413 1043360==    by 0xADBE27: AllocSetContextCreateInternal (aset.c:444)
==00:00:00:04.413 1043360==    by 0x762E31: PostmasterMain (postmaster.c:534)
==00:00:00:04.413 1043360==    by 0x5E9543: main (main.c:231)
==00:00:00:04.413 1043360==
...
==00:00:00:04.413 1043360==
==00:00:00:04.414 1043360== Exit program on first error (--exit-on-first-error=yes)
2026-04-12 08:03:29.349 EEST postmaster[1043298] LOG:  client backend (PID 1043360) exited with exit code 1
2026-04-12 08:03:29.349 EEST postmaster[1043298] DETAIL:  Failed process was running: SELECT test_saslprep('\xef');

The corresponding code:
    src_len = VARSIZE_ANY_EXHDR(string);
    src = VARDATA_ANY(string);

    /*
     * Copy the input given, to make SASLprep() act on a sanitized string.
     */
    input_data = palloc0(src_len + 1);
    strlcpy(input_data, src, src_len + 1);

That is, strlcpy() tries to evaluate strlen() for src, which contains only
one byte without null terminator.

skink tests this module successfully [1] by some reason:
================================== 151/394 ===================================
test:         test_saslprep - postgresql:test_saslprep/regress
start time:   13:26:00
duration:     8.04s
result:       exit status 0
command:      ...
----------------------------------- stdout -----------------------------------
# executing test in /home/bf/bf-build/skink-master/HEAD/pgsql.build/testrun/test_saslprep/regress group test_saslprep
test regress
# initializing database system by copying initdb template
# using temp instance on port 40096 with PID 3982971
ok 1         - test_saslprep                            1971 ms
1..1
# All 1 tests passed.
# test succeeded

[1] https://buildfarm.postgresql.org/cgi-bin/show_stage_log.pl?nm=skink&dt=2026-04-11%2011%3A31%3A01&stg=misc-check

Best regards,
Alexander