Re: Refactoring the checkpointer's fsync request queue

On Thu, Apr 04, 2019 at 05:39:14PM +1300, Thomas Munro wrote:
> On Thu, Apr 4, 2019 at 5:36 PM Andres Freund <andres@anarazel.de> wrote:
> > On 2019-04-03 21:19:45 -0700, Shawn Debnath wrote:
> > > +typedef struct FileTag
> > > +{
> > > +     int16           handler;                /* SyncRequstHandler value, saving space */
> > > +     int16           forknum;                /* ForkNumber, saving space */
> > > +     RelFileNode rnode;
> > > +     BlockNumber segno;
> > > +} FileTag;
> >
> > Seems odd to me to use BlockNumber for segno.
> 
> That is a tradition in md.c code.  I had a new typedef SegmentNumber
> in all sync.{c,h} stuff in an earlier version, but had trouble
> figuring out where to define it...

Thomas, this is why I had defined segment.h with the contents below :-)

+++ b/src/include/storage/segment.h
[...]
+/*
+ * Segment Number:
+ *
+ * Each relation and its forks are divided into segments. This
+ * definition formalizes the definition of the segment number.
+ */
+typedef uint32 SegmentNumber;
+
+#define InvalidSegmentNumber ((SegmentNumber) 0xFFFFFFFF)

My last iteration, v12, patch had it. See [1] for comments on removal of 
segment.h.


[1] 
https://www.postgresql.org/message-id/20190403214423.GA45392%40f01898859afd.ant.amazon.com

-- 
Shawn Debnath
Amazon Web Services (AWS)