SPI_exec ERROR in pl/r of R 3.2.4 on PostgreSQL on Windows 7 - Mailing list pgsql-hackers

From Andre Mikulec
Subject SPI_exec ERROR in pl/r of R 3.2.4 on PostgreSQL on Windows 7
Date
Msg-id CO2PR03MB2407927CD1C62BB0DA0E33BD9C660@CO2PR03MB2407.namprd03.prod.outlook.com
Whole thread Raw
Responses Re: SPI_exec ERROR in pl/r of R 3.2.4 on PostgreSQL on Windows 7  (Joe Conway <mail@joeconway.com>)
List pgsql-hackers

I am working with pl/r compiled for R.3.2.4 64 bit on PostgreSQL 9.5.1 64 bit on Windows 7 64 bit



At the end of this issue, I am getting the following error.
https://github.com/postgres-plr/plr/issues/1

  ERROR:  could not open file "base/12373/2663": No such file or directory
  LINE 1: SELECT NULL FROM pg_catalog.pg_class WHERE relname = 'plr_mo...
                         ^
QUERY:  SELECT NULL FROM pg_catalog.pg_class WHERE relname = 'plr_modules' AND relnamespace = 2200

The error seems to be coming from SPI_exec.

If I run this SQL manually from psql
SELECT NULL FROM pg_catalog.pg_class WHERE relname = 'plr_modules' AND relnamespace = 2200

The result is returned and is correct.

* The problem is not my hard disk. *
I am running multiple versions of PostgreSQL on the same hard disk. *

The following run fine.
Regular Windows pre-compiled PostgreSQL 9.4.1 downladed from downloaded from postgresql.org
Regular Windows pre-compiled PostgreSQL 9.5.1 downladed from downloaded from postgresql.org
Regular Windows pre-compiled PostgreSQL 9.5.2 downladed from downloaded from postgresql.org

THe problem is not security.
I am gave 'Full Access' to Administators group , EveryOne group, and Users group to
the directories containing all of the PostgreSQL directries containing both/either data and binaries.

I have shutdown all virus software: AVG.

The pl/r and plr.dll for R 3.1.2 64 bit runs fine on PostgreSQL 9.4.1 64bit on Windows 7 64
The pl/r source code has not changed at least since PostgreSQL 9.4.1.

I have physically examined the pl/r source code.
It seems relatively simple to understand.  

THe error seems to only come from here.
https://raw.githubusercontent.com/jconway/plr/master/plr.c


static bool
haveModulesTable(Oid nspOid)
{
StringInfo sql = makeStringInfo();
char   *sql_format = "SELECT NULL "
"FROM pg_catalog.pg_class "
"WHERE "
"relname = 'plr_modules' AND "
"relnamespace = %u";
    int  spiRc;

appendStringInfo(sql, sql_format, nspOid);

spiRc = SPI_exec(sql->data, 1);
if (spiRc != SPI_OK_SELECT)
/* internal error */
elog(ERROR, "haveModulesTable: select from pg_class failed");

return SPI_processed == 1;
}


I noticed that the using in the SPI_exec function *seems* to be similar in the source code.

https://raw.githubusercontent.com/postgres/postgres/3aff33aa687e47d52f453892498b30ac98a296af/src/test/regress/regress.c

query = (char *) palloc(100 + NAMEDATALEN * 3 +
strlen(fieldval) + strlen(fieldtype));

sprintf(query, "insert into %s select * from %s where %s = '%s'::%s",
SPI_getrelname(rel), SPI_getrelname(rel),
SPI_fname(tupdesc, 1),
fieldval, fieldtype);

if ((ret = SPI_exec(query, 0)) < 0)
elog(ERROR, "funny_dup17 (fired %s) on level %3d: SPI_exec (insert ...) returned %d",
when, *level, ret);



AND SPI_exec *seems* to be similar here

https://raw.githubusercontent.com/postgres/postgres/8b99edefcab1e82c43139a2c7dc06d31fb27b3e4/src/backend/commands/matview.c

StringInfoData querybuf;
initStringInfo(&querybuf);

/* Analyze the temp table with the new contents. */
appendStringInfo(&querybuf, "ANALYZE %s", tempname);
if (SPI_exec(querybuf.data, 0) != SPI_OK_UTILITY)
elog(ERROR, "SPI_exec failed: %s", querybuf.data);

It is defined here.

https://raw.githubusercontent.com/postgres/postgres/3aff33aa687e47d52f453892498b30ac98a296af/src/include/executor/spi.h

extern int SPI_execute(const char *src, bool read_only, long tcount);


https://raw.githubusercontent.com/postgres/postgres/39c283e498de1bb7c3d5beadfffcf3273ae8cc27/src/backend/executor/spi.c

/* Parse, plan, and execute a query string */
int
SPI_execute(const char *src, bool read_only, long tcount)
{
_SPI_plan plan;
int res;

if (src == NULL || tcount < 0)
return SPI_ERROR_ARGUMENT;

res = _SPI_begin_call(true);
if (res < 0)
return res;

memset(&plan, 0, sizeof(_SPI_plan));
plan.magic = _SPI_PLAN_MAGIC;
plan.cursor_options = 0;

_SPI_prepare_oneshot_plan(src, &plan);

res = _SPI_execute_plan(&plan, NULL,
InvalidSnapshot, InvalidSnapshot,
read_only, true, tcount);

_SPI_end_call(true);
return res;
}


/* Obsolete version of SPI_execute */
int
SPI_exec(const char *src, long tcount)
{
return SPI_execute(src, false, tcount);
}


My Big question is the following,

Has there been any change in the PostgreSQL IO code from 9.4.1. to 9.5.1
that may be possibly causing this problem ( in SPI_exec or elsewhere )?

  ERROR:  could not open file "base/12373/2663": No such file or directory

Any answers with any possibilities of any directions are welcome.

Thank you,
Andre Mikulec
Andre_Mikulec@Hotmail.com



pgsql-hackers by date:

Previous
From: Craig Ringer
Date:
Subject: Re: Timeline following for logical slots
Next
From: Petr Jelinek
Date:
Subject: Re: VS 2015 support in src/tools/msvc