On Mon, Sep 7, 2015 at 6:41 PM, Teodor Sigaev <teodor@sigaev.ru> wrote:
Some notices:
1) create-am.3.patch.gz As I understand, you didn't add schema name to access method. Why? Suppose, if we implement SQL-like interface for AM screation/dropping then we should provide a schema qualification for them
Per subsequent discussion it's unclear why we need to make access methods schema qualified.
2) create-am.3.patch.gz get_object_address_am() + switch (list_length(objname)) ... + case 2: + catalogname = strVal(linitial(objname)); + amname = strVal(lthird(objname)); ^^^^^^ seems, it should be lsecond()
Fixed.
3) create-am.3.patch.gz Suppose, RM_GENERIC_ID is part of generic-xlog.3.patch.gz
Fixed.
4) Makefile(s) As I can see, object files are lexicographically ordered
Fixed.
5) gindesc.c -> genericdesc.c in file header
Fixed.
6) generic-xlog.3.patch.gz I don't like an idea to split START_CRIT_SECTION and END_CRIT_SECTION to GenericXLogStart and GenericXLogFinish. User's code could throw a error or allocate memory between them and error will become a panic.
Fixed. Now, critical section is only inside GenericXLogFinish. GenericXLogRegister returns pointer to the page copies which could be modified. GenericXLogFinish swaps the data between page copy and page itself. That allow us to avoid critical section in used code.
------ Alexander Korotkov Postgres Professional: http://www.postgrespro.com The Russian Postgres Company