Unfortunately posix threading on SUN OS is not identical to Linux.
That was the idea behind Posix threads but there is enough wiggle room in the standard that porting is always a problem.
What's worse is I've never seen a good code porting guide or even a good book or guide about Posix threads on Linux. If any one knows of good ones please let me know



-- Sent from my HP Pre3


On Oct 17, 2013 20:15, Russell Beall <beall@usc.edu> wrote:

Hi,

I am trying to port a plugin from our Sun DS to 389. I worked through a number of API differences and got it to compile. When I tried to run a test I keep getting segmentation faults in the pthread library regardless of how I compile or link the code. I boiled it down to the following single line-item code which produces the crash:

int main (int argc, char **argv)
{
LDAP *ld = slapi_ldap_init( ldaphost, 389, 0, 1);
}

The stack trace is this:

#0 0x0000003c61009220 in pthread_mutex_lock () from /lib64/libpthread.so.0
#1 0x0000003dd3c240b9 in PR_Lock () from /lib64/libnspr4.so
#2 0x0000003dd70ae6b8 in set_snmp_interaction_row ()
from /usr/lib64/dirsrv/libslapd.so.0
#3 0x0000003dd7066771 in slapi_ldap_init_ext () from /usr/lib64/dirsrv/libslapd.so.0
#4 0x000000000040065f in main (argc=<value optimized out>,
argv=<value optimized out>) at basictest_crash.c:22

This is under RedHat 6 and this version of the directory:
389-Directory/1.2.11.23 B2013.275.1555

I had a much longer email here until I dug into the server code where the crash occurs. I found that when running the plugin code independently of the server, the snmp_collator code was not initialized and a thread mutex was still NULL when attempting to lock it. I hacked in this to force initialization:

diff ~/Downloads/389-ds-base-1.2.11.23/ldap/servers/slapd/snmp_collator.c ~/Downloads/389-ds-base-1.2.11.23-modified/ldap/servers/slapd/snmp_collator.c
239a240
> if (! interaction_table_mutex) interaction_table_mutex = PR_NewLock();

Linking against the hacked libslapd.so causes my test to function.

So… I've basically figured out what I needed to know by deep hacking, but I thought I should go ahead and post this in case it might be useful for discussion and possible patching.

When developing this plugin for Sun DS I worked with a locally executable version for testing as well as with it installed in the server. It would definitely be useful to continue being able to develop independently of the server without always having to plug it in to test.

Regards,
Russ.
--
389 users mailing list
389-users@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/389-users