rdelaage
/
postgresql
mirror of https://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Suppress -Wimplicit-fallthrough warning in new LIMIT WITH TIES code. 

The placement of the fall-through comment in this code appears not to
work to suppress the warning in recent gcc.  Move it to the bottom of
the case group, and add an assertion that we didn't get there through
some other code path.  Also improve wording of nearby comments.

Julien Rouhaud, comment hacking by me

Discussion: https://postgr.es/m/CAOBaU_aLdPGU5wCpaowNLF-Q8328iR7mj1yJAhMOVsdLwY+sdg@mail.gmail.com
This commit is contained in:
Tom Lane 2020-04-11 15:02:38 -04:00
parent 328c70997b
commit 35cb574aa8
1 changed files with 19 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
src/backend/executor/nodeLimit.c
View File

@ -138,17 +138,18 @@ ExecLimit(PlanState *pstate)
/*
* Forwards scan, so check for stepping off end of window. At
* the end of the window, the behavior depends on whether WITH
* TIES was specified: in that case, we need to change the
* state machine to LIMIT_WINDOWTIES. If not (nothing was
* specified, or ONLY was) return NULL without advancing the
* subplan or the position variable but change the state
* machine to record having done so
* TIES was specified: if so, we need to change the state
* machine to WINDOWEND_TIES, and fall through to the code for
* that case. If not (nothing was specified, or ONLY was)
* return NULL without advancing the subplan or the position
* variable, but change the state machine to record having
* done so.
*
* Once at the end, ideally, we can shut down parallel
* resources but that would destroy the parallel context which
* would be required for rescans. To do that, we need to find
* a way to pass down more information about whether rescans
* are possible.
* Once at the end, ideally, we would shut down parallel
* resources; but that would destroy the parallel context
* which might be required for rescans. To do that, we'll
* need to find a way to pass down more information about
* whether rescans are possible.
*/
if (!node->noCount &&
node->position - node->offset >= node->count)
@ -161,7 +162,7 @@ ExecLimit(PlanState *pstate)
else
{
node->lstate = LIMIT_WINDOWEND_TIES;
/* fall-through */
/* we'll fall through to the next case */
}
}
else
@ -177,8 +178,9 @@ ExecLimit(PlanState *pstate)
}
/*
* Tuple at limit is needed for comparation in subsequent
* execution to detect ties.
* If WITH TIES is active, and this is the last in-window
* tuple, save it to be used in subsequent WINDOWEND_TIES
* processing.
*/
if (node->limitOption == LIMIT_OPTION_WITH_TIES &&
node->position - node->offset == node->count - 1)
@ -194,7 +196,7 @@ ExecLimit(PlanState *pstate)
{
/*
* Backwards scan, so check for stepping off start of window.
* As above, change only state-machine status if so.
* As above, only change state-machine status if so.
*/
if (node->position <= node->offset + 1)
{
@ -213,6 +215,9 @@ ExecLimit(PlanState *pstate)
break;
}
Assert(node->lstate == LIMIT_WINDOWEND_TIES);
/* FALL THRU */
case LIMIT_WINDOWEND_TIES:
if (ScanDirectionIsForward(direction))
{