diff options
| author | Tom Lane | 2025-03-31 16:16:10 +0000 |
|---|---|---|
| committer | Tom Lane | 2025-03-31 16:16:32 +0000 |
| commit | 2fd3e2fa5c908d05ac4d8b6c2c80fd50c506f32e (patch) | |
| tree | b89f702348196d3f904984499b349d38ff42a26a | |
| parent | 0fcf02ad456862d8523a1918e2882e7697a52896 (diff) | |
Fix accidentally-harmless thinko in psqlscan_test_variable().
This code was passing literal strings to psqlscan_emit,
which is quite contrary to that function's specification:
"If you pass it something that is not part of the yytext
string, you are making a mistake". It accidentally worked
anyway, even in non-safe_encoding mode. psqlscan_emit
would compute a garbage "reference" pointer, but would
never dereference that since the passed string is all-ASCII.
So there's no live bug today, but that is a happenstance
outcome of psqlscan_emit's current implementation.
Let's make psqlscan_test_variable do what it's supposed to,
namely append directly to the output buffer. This is just
future-proofing against possible changes in psqlscan_emit,
so I don't feel a need to back-patch.
| -rw-r--r-- | src/fe_utils/psqlscan.l | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/src/fe_utils/psqlscan.l b/src/fe_utils/psqlscan.l index 4dbf2564276..fc5bfa597bb 100644 --- a/src/fe_utils/psqlscan.l +++ b/src/fe_utils/psqlscan.l @@ -1626,11 +1626,11 @@ psqlscan_test_variable(PsqlScanState state, const char *txt, int len) if (value != NULL) { - psqlscan_emit(state, "TRUE", 4); + appendPQExpBufferStr(state->output_buf, "TRUE"); free(value); } else { - psqlscan_emit(state, "FALSE", 5); + appendPQExpBufferStr(state->output_buf, "FALSE"); } } |