diff options
author | Tom Lane <tgl@sss.pgh.pa.us> | 2024-11-25 18:08:58 -0500 |
---|---|---|
committer | Tom Lane <tgl@sss.pgh.pa.us> | 2024-11-25 18:09:09 -0500 |
commit | 5b8728cd7f9d3d93b6ff9b48887084fdf0a46e4f (patch) | |
tree | b62eee91b8fa0d602753987f1a1c00aa0b7bb01f /src/backend/executor | |
parent | 4ba84de459532a7a79616e756f8bfd45ba4b8048 (diff) | |
download | postgresql-5b8728cd7f9d3d93b6ff9b48887084fdf0a46e4f.tar.gz postgresql-5b8728cd7f9d3d93b6ff9b48887084fdf0a46e4f.zip |
Fix NULLIF()'s handling of read-write expanded objects.
If passed a read-write expanded object pointer, the EEOP_NULLIF
code would hand that same pointer to the equality function
and then (unless equality was reported) also return the same
pointer as its value. This is no good, because a function that
receives a read-write expanded object pointer is fully entitled
to scribble on or even delete the object, thus corrupting the
NULLIF output. (This problem is likely unobservable with the
equality functions provided in core Postgres, but it's easy to
demonstrate with one coded in plpgsql.)
To fix, make sure the pointer passed to the equality function
is read-only. We can still return the original read-write
pointer as the NULLIF result, allowing optimization of later
operations.
Per bug #18722 from Alexander Lakhin. This has been wrong
since we invented expanded objects, so back-patch to all
supported branches.
Discussion: https://postgr.es/m/18722-fd9e645448cc78b4@postgresql.org
Diffstat (limited to 'src/backend/executor')
-rw-r--r-- | src/backend/executor/execExpr.c | 8 | ||||
-rw-r--r-- | src/backend/executor/execExprInterp.c | 14 |
2 files changed, 21 insertions, 1 deletions
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c index 8f7a5340059..69d36f70b3c 100644 --- a/src/backend/executor/execExpr.c +++ b/src/backend/executor/execExpr.c @@ -1192,6 +1192,14 @@ ExecInitExprRec(Expr *node, ExprState *state, state); /* + * If first argument is of varlena type, we'll need to ensure + * that the value passed to the comparison function is a + * read-only pointer. + */ + scratch.d.func.make_ro = + (get_typlen(exprType((Node *) linitial(op->args))) == -1); + + /* * Change opcode of call instruction to EEOP_NULLIF. * * XXX: historically we've not called the function usage diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c index 30c5a19aad6..42da9621789 100644 --- a/src/backend/executor/execExprInterp.c +++ b/src/backend/executor/execExprInterp.c @@ -1308,12 +1308,24 @@ ExecInterpExpr(ExprState *state, ExprContext *econtext, bool *isnull) * The arguments are already evaluated into fcinfo->args. */ FunctionCallInfo fcinfo = op->d.func.fcinfo_data; + Datum save_arg0 = fcinfo->args[0].value; /* if either argument is NULL they can't be equal */ if (!fcinfo->args[0].isnull && !fcinfo->args[1].isnull) { Datum result; + /* + * If first argument is of varlena type, it might be an + * expanded datum. We need to ensure that the value passed to + * the comparison function is a read-only pointer. However, + * if we end by returning the first argument, that will be the + * original read-write pointer if it was read-write. + */ + if (op->d.func.make_ro) + fcinfo->args[0].value = + MakeExpandedObjectReadOnlyInternal(save_arg0); + fcinfo->isnull = false; result = op->d.func.fn_addr(fcinfo); @@ -1328,7 +1340,7 @@ ExecInterpExpr(ExprState *state, ExprContext *econtext, bool *isnull) } /* Arguments aren't equal, so return the first one */ - *op->resvalue = fcinfo->args[0].value; + *op->resvalue = save_arg0; *op->resnull = fcinfo->args[0].isnull; EEO_NEXT(); |