On Wed, Dec 22, 2021 9:55 PM Amit Kapila <amit.kapil...@gmail.com> wrote: > On Wed, Dec 22, 2021 at 6:22 PM Amit Kapila <amit.kapil...@gmail.com> > wrote: > > > > On Wed, Dec 22, 2021 at 5:39 PM houzj.f...@fujitsu.com > > <houzj.f...@fujitsu.com> wrote: > > > > > > > > > 2) > > > +#include "utils/rel.h" > > > +#include "utils/lsyscache.h" > > > +#include "utils/memutils.h" > > > > > > It might be better to keep the header file in alphabetical order. > > > : > > > +#include "utils/lsyscache.h" > > > +#include "utils/memutils.h" > > > +#include "utils/rel.h" > > > > > > > Right, I'll take care of this as I am already making some other edits > > in the patch. > > > > Fixed this and made a few other changes in the patch that includes (a) passed > down the num_index_scans information in parallel APIs based on which it can > make the decision whether to reinitialize DSM or consider conditional parallel > vacuum clean up; (b) get rid of first-time variable in ParallelVacuumState as > that > is not required if we have num_index_scans information; (c) there seems to be > quite a few unnecessary includes in vacuumparallel.c which I have removed; (d) > unnecessary error callback info was being set in ParallelVacuumState in leader > backend; (e) changed/added comments at quite a few places. > > Can you please once verify the changes in the attached?
The changes look ok to me. I tested the patch for multi-pass parallel vacuum cases and ran 'make check-world', all the tests passed Best regards, Hou zj
