New/old style trigger API
The docs seem a little sketchy but the source implies that there is
a difference between new and old style triggers. I can't seem to find
this difference. I tried following the only current example I could
find but I get a core dump. Here is the backtrace.
#0 0x0 in ?? ()
#1 0x8136052 in fmgr_oldstyle (fcinfo=0xbfbfd134) at fmgr.c:433
#2 0x80b6b2d in ExecCallTriggerFunc (trigger=0x82ce030, trigdata=0xbfbfd1c8,
per_tuple_context=0x827acf0) at trigger.c:865
#3 0x80b6e15 in ExecBRUpdateTriggers (estate=0x82fbf50, tupleid=0xbfbfd264,
newtuple=0x83094a8) at trigger.c:1008
#4 0x80bcf3a in ExecReplace (slot=0x8308018, tupleid=0xbfbfd264,
estate=0x82fbf50) at execMain.c:1416
#5 0x80bcc82 in ExecutePlan (estate=0x82fbf50, plan=0x82fbec8,
operation=CMD_UPDATE, numberTuples=0, direction=ForwardScanDirection,
destfunc=0x83092f0) at execMain.c:1127
#6 0x80bc27b in ExecutorRun (queryDesc=0x8307cc0, estate=0x82fbf50,
feature=3, count=0) at execMain.c:233
#7 0x80ff29b in ProcessQuery (parsetree=0x82f6e20, plan=0x82fbec8,
dest=Remote) at pquery.c:277
#8 0x80fddd3 in pg_exec_query_string (
query_string=0x82f6030 "UPDATE bgroup SET actypid = 1,ddate = NULL,edate = '2001-07-23',mail = '',ftp = '',dns = 'f',bname = 'jpsantos',bgdesc = '',isp = 'VEX',pdate = '2001-04-23',pmon = 5,bgroup_active = 't',sdate = '1999-"...,
dest=Remote, parse_context=0x827a720) at postgres.c:808
#9 0x80fedd1 in PostgresMain (argc=4, argv=0xbfbfd4c8, real_argc=9,
real_argv=0xbfbfdc54, username=0x826325d "darcy") at postgres.c:1905
#10 0x80ea2ad in DoBackend (port=0x8263000) at postmaster.c:2114
#11 0x80e9eb2 in BackendStartup (port=0x8263000) at postmaster.c:1897
#12 0x80e91aa in ServerLoop () at postmaster.c:995
#13 0x80e8b7c in PostmasterMain (argc=9, argv=0xbfbfdc54) at postmaster.c:685
#14 0x80caff4 in main (argc=9, argv=0xbfbfdc54) at main.c:175
#15 0x806bc81 in ___start ()
It gets to fmgr_oldstyle() and then dies from a jump to a null pointer.
Can someone please tell me how to make my function a newstyle one. I
will update the docs once I have it working.
--
D'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves
http://www.druid.net/darcy/ | and a sheep voting on
+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.
darcy@druid.net (D'Arcy J.M. Cain) writes:
It gets to fmgr_oldstyle() and then dies from a jump to a null pointer.
Can someone please tell me how to make my function a newstyle one.
Perhaps you forgot the PG_FUNCTION_INFO_V1 declaration? See
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/trigger-examples.html
regards, tom lane
Thus spake Tom Lane
darcy@druid.net (D'Arcy J.M. Cain) writes:
It gets to fmgr_oldstyle() and then dies from a jump to a null pointer.
Can someone please tell me how to make my function a newstyle one.Perhaps you forgot the PG_FUNCTION_INFO_V1 declaration? See
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/trigger-examples.html
Yah, I was looking at chapter 22 on SPI programming. I assume that the
same should apply there. Shall I go ahead and add it to the docs in that
chapter as well?
I wonder if there is a relation to another problem I am having. On AIX
I can compile my chkpass function (in contrib which I just updated) OK
but when I use it I get a similar core dump there that I don't see on
NetBSD. Does it require the same interface? Shall I update those pages
as well?
BTW, here is the link command I needed to get as far as I did.
ld -G -o $@ $< -bexpall -bnoentry -lc
--
D'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves
http://www.druid.net/darcy/ | and a sheep voting on
+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.darcy@druid.net (D'Arcy J.M. Cain) writes:
Yah, I was looking at chapter 22 on SPI programming. I assume that the
same should apply there. Shall I go ahead and add it to the docs in that
chapter as well?
Uh ... where? I see nothing related to trigger programming in chapter 22.
There's an example of an old-style function at
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/spi-examples.html
but it's not a trigger.
regards, tom lane
Thus spake Tom Lane
darcy@druid.net (D'Arcy J.M. Cain) writes:
Yah, I was looking at chapter 22 on SPI programming. I assume that the
same should apply there. Shall I go ahead and add it to the docs in that
chapter as well?Uh ... where? I see nothing related to trigger programming in chapter 22.
There's an example of an old-style function at
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/spi-examples.html
but it's not a trigger.
Ah. So PG_FUNCTION_INFO_V1() is strictly for triggers then. The name
suggested it was a little more global than that.
--
D'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves
http://www.druid.net/darcy/ | and a sheep voting on
+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.darcy@druid.net (D'Arcy J.M. Cain) writes:
There's an example of an old-style function at
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/spi-examples.html
but it's not a trigger.
Ah. So PG_FUNCTION_INFO_V1() is strictly for triggers then.
No, it's for new-style functions. See
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/xfunc-c.html
regards, tom lane
Thus spake Tom Lane
darcy@druid.net (D'Arcy J.M. Cain) writes:
There's an example of an old-style function at
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/spi-examples.html
but it's not a trigger.Ah. So PG_FUNCTION_INFO_V1() is strictly for triggers then.
No, it's for new-style functions. See
http://www.ca.postgresql.org/users-lounge/docs/7.1/postgres/xfunc-c.html
I'm so confused. :-)
OK, so all functions can be upgraded to version 1 calling sequence, not
just ones used for triggers. I think I have it but the docs could still
use some work. If someone wants to check out the latest version of my
chkpass function in contrib and let me know if I seem to have the right
idea now then I will go ahead and try to improve the docs.
--
D'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves
http://www.druid.net/darcy/ | and a sheep voting on
+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.darcy@druid.net (D'Arcy J.M. Cain) writes:
OK, so all functions can be upgraded to version 1 calling sequence, not
just ones used for triggers.
Now you're getting there. Actually triggers *must* be upgraded to the
new calling sequence, because we don't support old-style triggers
anymore. But for other functions it's optional.
regards, tom lane
Thus spake Tom Lane
darcy@druid.net (D'Arcy J.M. Cain) writes:
OK, so all functions can be upgraded to version 1 calling sequence, not
just ones used for triggers.Now you're getting there. Actually triggers *must* be upgraded to the
new calling sequence, because we don't support old-style triggers
anymore. But for other functions it's optional.
OK, I am going to try to find time to update the other pages. In particular
I will update the examples and add the list of macros.
--
D'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves
http://www.druid.net/darcy/ | and a sheep voting on
+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.