[kritzingerf]

Good day all,

I joined this list in the hope of getting some assistance with Pervasive BTRIEVE.

The background:
I need an application that will connect to a DB2 database, and export certain tables to brieve files. The application was originally written in C on OS/2. I have now ported it to Linux using C and Pervasive Developer engine 9.50.077.002

Program Segmentation faults in first BTRV/BTRVID call...

Here is a screen dump of the Stack Trace...

Checking Btrieve Versions...

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread -1208097088 (LWP 28311)]
0x08eae700 in ?? ()
(gdb) bt
#0 0x08eae700 in ?? ()
#1 0x002e40e1 in __dynamic_cast () from /usr/lib/libstdc++-libc6.1-1.so.2
#2 0x004e40cb in ps::text::charset::ByteToCharConverter::getConvert er () from /usr/local/psql/lib/libpscl.so.2
#3 0x004e0252 in ps::String::internalSetTo () from /usr/local/psql/lib/libpscl.so.2
#4 0x004dbed9 in ps::String::String () from /usr/local/psql/lib/libpscl.so.2
#5 0x001627f0 in MkdeidAdmin () from /usr/local/psql/lib/libpsqlmif.so.9
#6 0x00162757 in MkdeidAdmin () from /usr/local/psql/lib/libpsqlmif.so.9
#7 0x00151311 in DBUSetInfo () from /usr/local/psql/lib/libpsqlmif.so.9
#8 0x0014d77f in mkdeID () from /usr/local/psql/lib/libpsqlmif.so.9
#9 0x0014da04 in mkdeUTF8PathID () from /usr/local/psql/lib/libpsqlmif.so.9
#10 0x00149058 in BTRCALLID () from /usr/local/psql/lib/libpsqlmif.so.9
#11 0x00148c93 in BTRCALLID () from /usr/local/psql/lib/libpsqlmif.so.9
#12 0x00147557 in BTRVID () from /usr/local/psql/lib/libpsqlmif.so.9
#13 0x08049cbd in btrvCheckStatus () at sql2btrv.c:659
#14 0x08049dc4 in main (argc=1, argv=0xbff4ade4) at sql2btrv.c:726



Any advice / guidance will be highly valued!

Thanx

Francois

[mirtheil]

A few questions:
1. What Linux distribution are you using?
2. What is the actual code that's failing?
3. Does the Btrieve sample (included in the Btrieve SDK) compile and run?

[kritzingerf]

Red Hat ES

Btrieve sample compiles and run, also test database work fine....

Here is the actual code in C:

void btrvCheckStatus()
{
BTI_BYTE btrieveLoaded = FALSE;
BTI_BYTE posBlock1[128];
VERSION_STRUCT versionBuffer[3];
BTI_WORD dataLen = 0;
BTI_BYTE keyBuf1[255];
BTI_WORD keyNum = 0;
BTI_SINT status;
BTI_SINT i;

memset(versionBuffer, 0, sizeof(versionBuffer));
dataLen = sizeof(versionBuffer);

printf("Checking Btrieve Versions...\r\n");
status = BTRVID(B_VERSION,
posBlock1,
&versionBuffer,
&dataLen,
keyBuf1,
keyNum,
(BTI_BUFFER_PTR)&clientID);

if (status == B_NO_ERROR) {
printf("Btrieve Versions returned are: \r\n");
for (i = 0; i < 3; i++) {
if (versionBuffer[i].Version != 0) {
printf(" %d.%d %c\r\n", versionBuffer[i].Version,
versionBuffer[i].Revision,
versionBuffer[i].MKDEId);
}
}
printf("\n");
btrieveLoaded = TRUE;
} else {
printf("Btrieve B_VERSION status = %d\n", status);
if (status != B_RECORD_MANAGER_INACTIVE) {
btrieveLoaded = TRUE;
}
}
}


The GPF occurs in the bold BTRVID call....

Thanx and cheers

F.

[mirtheil]

If the sample code compiles and runs, then the problem is most likely in your code. One thing I notice is that you don't initialize the clientID variable.
Specifically, this code is used in the sample:

memset(clientID.networkAndNode, 0, sizeof(clientID.networkAndNode));
  memcpy(clientID.applicationID, "MT", 2);  /* must be greater than "AA" */
  clientID.threadID = MY_THREAD_ID;

[kritzingerf]

Hi Mirtheil,

clientId is defined as global, and initialised in the main exactly as the sample shows. I have in GDB checked that clientId actually contains all appropriate values as set in the sample.

The only thing I can think is that somehow the environment believes we are running in a full client/server setup, which I am not. My development box is where the the SDK is installed. I want to run in single machine mode with direct file access. But I am at a loss on how to check that....

thanx

Francois

[mirtheil]

The interface / engine doesn't distinguish between the client/server setup and the standalone setup. At the point your application crashes, it doesn't matter when engine/client is installed.
Normally, when an application crashes, it's because of an uninitialized variable.
Just as a test, what happens if you initialize the ClientID structure right before the Version call?