verifychecksum: In case of checksum mismatch, expected and actual are reversed in error message

[mbanck]

No description provided.

[mbanck]

patch for this:

--- verifychecksum.c.orig       2018-01-23 14:12:23.000000000 +0100
+++ verifychecksum.c.new        2018-01-23 14:35:30.000000000 +0100
@@ -23,7 +23,7 @@
   if (phdr->pd_checksum != checksum)
     {
       printf("%s: blkno %d, expected %x, found %x\n",
-            filepath, blkno, checksum, phdr->pd_checksum);
+            filepath, blkno, phdr->pd_checksum, checksum);
       return FALSE;
     }

[snaga]

In this case, phdr->pd_checksum is a stored value in the page header and checksum is a calculated (on-the-fly) value from the page content. Here, the latter one should be treated as the correct checksum.

So, I handle phdr->pd_checksum as found, and checksum as expected.

Doesn't it make sense?

[mbanck]

I was looking at it from Postgres' perspective: it has stored phdr->pd_checksum in the page header, so it expects this checksum to be there. But actually, we found checksum.

[snaga]

I see.
Need more time to look for more appropriate words.

[snaga]

How about "page header" for phdr->pd_checksum and "calculated" as checksum?
Does it make sense?

[mbanck]

Sounds good to me, yeah.

[snaga]

Committed. 9ef3e48

[snaga]

Merged. #60