Failed regression tests
I recently compiled and installed PostgreSQL 7.0 on a P-III workstation
running SuSE Linux 6.2 (2.2.10 kernel, gcc ver. 2.7.2.3). The compilation
completed with some non-fatal warnings, but when I ran the sequential
regression tests, five failures occurred.
The first two failures may reflect a 'formating' problem:
1. int8, test 13: a "$" appears left-justified in all entries in 2nd col
2. numeric, test 13: a "$" appears left-justified in all entries in 2nd col
I am using an international code page (LANG=en_US.iso88591), and perhaps the
$ appears because I used /.configure --enable-locale?
The next two failures may reflect a 'sorting' problem:
3. select_implicit: output lines sorted A,B,b,C,c,X instead of A,B,C,X,b,c
4. select_having: output lines sorted b,X instead of X,b
The final "select_view" failure is rather messy:
*** expected/select_views.out Sat Jan 8 22:48:37 2000
--- results/select_views.out Fri May 26 13:25:43 2000
***************
*** 415,420 ****
--- 415,434 ----
I- 580 | 21
I- 580 | 22
I- 580 | 22
+ I- 580/I-680 Ramp | 2
+ I- 580/I-680 Ramp | 2
+ I- 580/I-680 Ramp | 2
+ I- 580/I-680 Ramp | 2
+ I- 580/I-680 Ramp | 2
+ I- 580/I-680 Ramp | 2
+ I- 580/I-680 Ramp | 4
+ I- 580/I-680 Ramp | 4
+ I- 580/I-680 Ramp | 4
+ I- 580/I-680 Ramp | 4
+ I- 580/I-680 Ramp | 5
+ I- 580/I-680 Ramp | 6
+ I- 580/I-680 Ramp | 6
+ I- 580/I-680 Ramp | 6
I- 580 Ramp | 2
I- 580 Ramp | 2
I- 580 Ramp | 2
***************
*** 665,684 ****
I- 580 Ramp | 8
I- 580 Ramp | 8
I- 580 Ramp | 8
- I- 580/I-680 Ramp | 2
- I- 580/I-680 Ramp | 2
- I- 580/I-680 Ramp | 2
- I- 580/I-680 Ramp | 2
- I- 580/I-680 Ramp | 2
- I- 580/I-680 Ramp | 2
- I- 580/I-680 Ramp | 4
- I- 580/I-680 Ramp | 4
- I- 580/I-680 Ramp | 4
- I- 580/I-680 Ramp | 4
- I- 580/I-680 Ramp | 5
- I- 580/I-680 Ramp | 6
- I- 580/I-680 Ramp | 6
- I- 580/I-680 Ramp | 6
I- 680 | 2
I- 680 | 2
I- 680 | 2
--- 679,684 ----
My config.log reports 3 problems which may have caused the failures:
-----
configure: In function `main':
configure:4885: storage size of `semun' isn't known
configure: failed program was:
#line 4880 "configure"
#include "confdefs.h"
#include <sys/types.h>
#include <sys/ipc.h>
#include <sys/sem.h>
int main() {
union semun semun;
; return 0; }
-----
configure:4904: checking for fcntl(F_SETLK)
configure:4916: gcc -o conftest -O2 conftest.c -lcrypt -lnsl -ldl -lm -lreadconfigure: In function `main':
configure:4910: `SEEK_SET' undeclared (first use in this function)
configure:4910: (Each undeclared identifier is reported only once
configure:4910: for each function it appears in.)
configure: failed program was:
#line 4906 "configure"
#include "confdefs.h"
#include <fcntl.h>
int main() {
struct flock lck;
lck.l_whence = SEEK_SET; lck.l_start = lck.l_len = 0;
lck.l_type = F_WRLCK;
fcntl(0, F_SETLK, &lck);
; return 0; }
------
configure:6446: gcc -o conftest -O2 conftest.c -lcrypt -lnsl -ldl -lm -lread
cconfigure: failed program was:
#line 6418 "configure"
#include "confdefs.h"
typedef long int int64;
/* These are globals to discourage the compiler from folding all the
* arithmetic tests down to compile-time constants.
*/
int64 a = 20000001;
int64 b = 40000005;
int does_int64_work()
{
int64 c,d;
if (sizeof(int64) != 8)
return 0; /* doesn't look like the right size */
/* Do perfunctory checks to see if 64-bit arithmetic seems to work */
c = a * b;
d = (c + b) / b;
if (d != a+1)
return 0;
return 1;
}
main() {
exit(! does_int64_work());
}
Finally, the /.configure output included a dozen or so check ... (no)
entries (including some header files), which may contribute to the failures.
Have I banged into configure bugs, missing or misplaced Linux libraries, or
my own stupidity?
Thanks for any pointers,
--
Bruce Saunders
b.saunders@mindspring.com writes:
I recently compiled and installed PostgreSQL 7.0 on a P-III workstation
running SuSE Linux 6.2 (2.2.10 kernel, gcc ver. 2.7.2.3). The compilation
completed with some non-fatal warnings, but when I ran the sequential
regression tests, five failures occurred.
I am using an international code page (LANG=en_US.iso88591), and perhaps the
$ appears because I used /.configure --enable-locale?
Offhand all of those look like they might be due to the LANG setting.
Try restarting the postmaster with LANG unset and then rerun the
regression tests to see what you get.
regards, tom lane