sqgfault on initdb with current CVS - Mailing list pgsql-hackers

From Ross J. Reedstrom
Subject sqgfault on initdb with current CVS
Date
Msg-id 20000306183722.A7703@rice.edu
Whole thread Raw
Responses Re: [HACKERS] sqgfault on initdb with current CVS
List pgsql-hackers
Sorry about the length of this one, but I wanted to include backtrace
info.  I've been getting a segfault when I try to setup a test version
of the current CVS tree, to run along beside a 'production' 6.5.X server.

This is how I've configured it:

wallace$./configure --prefix=/usr/local/pgsql7.0/ --with-pgport=6666 --enable-debug

After which make; make install works fine.

On my machine at home, where I can let pgsql take the default location
and port, I've had no trouble. With this one, at work, I get a segfault
when initdb tries to execute postgres in bootstrap mode. I've managed
to run postgres under gdb and capture a backtrace:

wallace$ which initdb
/usr/local/pgsql7.0/bin/initdb
wallace$ export | grep -i PG
declare -x LD_LIBRARY_PATH="/usr/local/pgsql7.0/lib"
declare -x PGDATA="/usr/local/pgsql7.0/data"
declare -x PGLIB="/usr/local/pgsql7.0/lib"
declare -x PWD="/extra/src/pgsql-current/src"
{PATH deleted}

{already ran initdb -d -n, then, after it failed, deleted
${PGDATA}/pg_control, and created an already sed processed version of
the template (just subs for PGUID) }

wallace$ gdb postgres  
GNU gdb 19990928
<deleted banner>
This GDB was configured as "i686-pc-linux-gnu"...
(gdb) set args -boot -x -C -F -D/usr/local/pgsql7.0/data -d template1 </home/reedstrm/template1.sql
(gdb) run
Starting program: /usr/local/pgsql7.0/bin/postgres -boot -x -C -F -D/usr/local/pgsql7.0/data -d template1
</home/reedstrm/template1.sql
<proname name> 
<proowner int4> 
<prolang oid> 
<proisinh bool> 
<proistrusted bool> 
<proiscachable bool> 
<pronargs int2> 
<proretset bool> 
<prorettype oid> 
<proargtypes oidvector> 
<probyte_pct int4> 
<properbyte_cpu int4> 
<propercall_cpu int4> 
<prooutin_ratio int4> 
<prosrc text> 
<probin bytea> 

> creating bootstrap relation

Program received signal SIGSEGV, Segmentation fault.
0x80eb266 in filepath (filename=0x81b2260 "pg_proc") at fd.c:529

529                     len = strlen(DatabasePath) + strlen(filename) + 2;
(gdb) bt
#0  0x80eb26
#1  0x80eb456 in FileNameOpenFile (fileName=0x81b2260 "pg_proc", fileFlags=194, fileMode=384)   at fd.c:657
#2  0x80f1d44 in mdcreate (reln=0x81b18b0) at md.c:128
#3  0x80f2da7 in smgrcreate (which=0, reln=0x81b18b0) at smgr.c:134
#4  0x8081a9d in heap_storage_create (rel=0x81b18b0) at heap.c:352
#5  0x8081a4c in heap_create (relname=0x81b8d48 "pg_proc", tupDesc=0x81bdd28, isnoname=0 '\000',    istemp=0,
storage_create=1'\001') at heap.c:329
 
#6  0x807e95c in Int_yyparse () at bootparse.y:168
#7  0x8080557 in BootstrapMain (argc=7, argv=0xbffffc38) at bootstrap.c:395
#8  0x80afcee in main (argc=8, argv=0xbffffc34) at main.c:100

{grovel around in fd.c: looks like DatabasePath should be set by
SetDatabasePath, which uses ExpandDatabasePath to the value of DataDir +
dbName (template1) but somehow we missed it}

(gdb) print DatabasePath
$1 = 0x0
(gdb) print DataDir     
$2 = 0xbffffd97 "/usr/local/pgsql7.0/data"
(gdb) exit

Anyone have any ideas? I can recreate at will.

Ross
-- 
Ross J. Reedstrom, Ph.D., <reedstrm@rice.edu> 
NSBRI Research Scientist/Programmer
Computer and Information Technology Institute
Rice University, 6100 S. Main St.,  Houston, TX 77005


pgsql-hackers by date:

Previous
From: Thomas Lockhart
Date:
Subject: Re: [HACKERS] TransactionStateData and AbsoluteTime
Next
From: Tatsuo Ishii
Date:
Subject: Re: [HACKERS] sqgfault on initdb with current CVS