Re: some linker troubles with rc5 on sun studio 9 ... - Mailing list pgsql-hackers

From Hans-Jürgen Schönig
Subject Re: some linker troubles with rc5 on sun studio 9 ...
Date
Msg-id 41FB59B5.4000205@cybertec.at
Whole thread Raw
In response to Re: some linker troubles with rc5 on sun studio 9 ...  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Tom Lane wrote:
> Hans-Jürgen Schönig <postgres@cybertec.at> writes:
> 
>>"tuptoaster.c", line 966: member can not have variably modified type: data
> 
> 
> We've seen that before.  Apparently there are some versions of Sun's
> compiler that are too stupid to reduce this constant expression to a
> constant.  File a bug with Sun.
> 
>             regards, tom lane
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 8: explain analyze is your friend
> 


As Tom pointed it this is truly a compiler bug of zthe compiler included 
in Sun Studio 9. It seems as if Sun is attempting to fix that.

Here is a test case which will fail on Sun CC 5.6.

#include <stddef.h>
#include <stdio.h>

typedef struct {
int s1;
char data[10];
unsigned char bits[1];
} Ss;

int main(int argc, char **argv){
struct {
int ap;
char data[offsetof(Ss,bits)];
}s1;

printf("offsetof: %d, sizeof %d\n",
offsetof(Ss, bits),
sizeof(s1.data));
return 0;
}


When I'm using Studio 9 cc, compilation failed:>/opt/F9/SUNWspro/bin/cc -V -Xa -o test test.c
cc: Sun C 5.6 2004/07/15
acomp: Sun C 5.6 2004/07/15
"test.c", line 14: member can not have variably modified type: data
cc: acomp failed for test.c


8.0.0 compiles now with --disable-spinlocks.

When spinlocks are enabled we will see one more problem ...

gmake[4]: Entering directory 
`/usr/share/src/pg/postgresql-8.0.0/src/backend/storage/lmgr'
/opt/SUNWspro/bin/cc -Xa -v -g -I../../../../src/include -c -o lmgr.o lmgr.c
/opt/SUNWspro/bin/cc -Xa -v -g -I../../../../src/include -c -o lock.o lock.c
/opt/SUNWspro/bin/cc -Xa -v -g -I../../../../src/include -c -o proc.o proc.c
/opt/SUNWspro/bin/cc -Xa -v -g -I../../../../src/include -c -o 
deadlock.o deadlock.c
/opt/SUNWspro/bin/cc -Xa -v -g -I../../../../src/include -c -o lwlock.o 
lwlock.c
/opt/SUNWspro/bin/cc -Xa -v -g -I../../../../src/include -c -o spin.o spin.c
/opt/SUNWspro/bin/cc -Xa -v -g -I../../../../src/include -c -o s_lock.o 
s_lock.c
/opt/SUNWspro/prod/bin/fbe: "/tmp/yabeAAAQ.aaFl", line 277: error: can't 
compute difference between symbols in different segments
Failure in /opt/SUNWspro/prod/bin/fbe, status = 0x100
Fatal Error exec'ing /opt/SUNWspro/prod/bin/fbe
cc: acomp failed for s_lock.c
gmake[4]: *** [s_lock.o] Error 2
gmake[4]: Leaving directory 
`/usr/share/src/pg/postgresql-8.0.0/src/backend/storage/lmgr'
gmake[3]: *** [lmgr-recursive] Error 2
gmake[3]: Leaving directory 
`/usr/share/src/pg/postgresql-8.0.0/src/backend/storage'
gmake[2]: *** [storage-recursive] Error 2
gmake[2]: Leaving directory `/usr/share/src/pg/postgresql-8.0.0/src/backend'
gmake[1]: *** [all] Error 2
gmake[1]: Leaving directory `/usr/share/src/pg/postgresql-8.0.0/src'
gmake: *** [all] Error 2


The nice thing is: This error is not documented and we have not found a 
flag to preserve /tmp/yabeAAAQ.aaFl (which makes this error very useful).

google does not contain information about this issue because all four 
sites I have found are down.

Has anybody ever seen something like that before?
Hans

-- 
Cybertec Geschwinde u Schoenig
Schoengrabern 134, A-2020 Hollabrunn, Austria
Tel: +43/660/816 40 77
www.cybertec.at, www.postgresql.at



pgsql-hackers by date:

Previous
From: Oleg Bartunov
Date:
Subject: Re: strange 'vacuum verbose analyze' behaviour
Next
From: Mischa
Date:
Subject: Re: Group-count estimation statistics