On Thu, Jan 12, 2023 at 10:44:33AM -0800, Nathan Bossart wrote:
> There's another "return" later on in this PG_TRY block. I wonder if it's
> possible to detect this sort of thing at compile time.
Note also:
src/pl/tcl/pltcl.c- * PG_CATCH();
src/pl/tcl/pltcl.c- * {
src/pl/tcl/pltcl.c- * pltcl_subtrans_abort(interp, oldcontext, oldowner);
src/pl/tcl/pltcl.c- * return TCL_ERROR;
src/pl/tcl/pltcl.c- * }
This is documented once, repeated twice:
src/pl/plpython/plpy_spi.c- * PG_CATCH();
src/pl/plpython/plpy_spi.c- * {
src/pl/plpython/plpy_spi.c- * <do cleanup>
src/pl/plpython/plpy_spi.c- * PLy_spi_subtransaction_abort(oldcontext, oldowner);
src/pl/plpython/plpy_spi.c- * return NULL;
src/pl/plpython/plpy_spi.c- * }
I don't quite get why this would be a sane thing to do here when
aborting a subtransaction in pl/python, but my experience with this
code is limited.
--
Michael