| Title: | Digital PATHWORKS 32 |
| Moderator: | SPELNK::curless |
| Created: | Fri Nov 01 1996 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 337 |
| Total number of notes: | 1612 |
One of our customer migrated his self written application from PW4.1 to PW32 on NT. He converted the code to use the SKT functions: previously (if I understood well) they converted the Microsoft static library to a DLL what they imported into the Borlan compiler, and this worked well. Now after they converted the source to use the Skt DLLs: for example the SktDnetcon function with new arguments and return values, they observed that the NT loops at 100% CPU consumption during the connection time or during the timeout value they give as argument to the Dnetconn function call if the node is not reachable. Is there an explanation for this or is this a bug? Since I am not quite familiar with DECNET programming, can someone give me advice on this? The customer is very used with such environments and reported already several problems in previous Pathworks DECNET programming V5 and received patches for that it... Jean-Paul, (Brussels)
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 223.1 | JAMIN::WASSER | John A. Wasser | Mon Apr 14 1997 14:22 | 15 | |
> NT loops at 100% CPU consumption during the connection time or during the > timeout value they give as argument to the Dnetconn function call > Is there an explanation for this or is this a bug? It is a bug in the PATHWORKS 32 DECnet stack. No fix is currently planned for V7.0A. They can work around the bug by doing only non-blocking operations. Since SktDnetcon() creates a socket and then does a blocking SktConnect() you have no choice but to abandon SktDnetcon() and call SktConnect() directly. Also make sure that this problem gets escalated so that it will eventually get fixed. | |||||
| 223.2 | BRSSWS::BERGMANS | Tue Apr 15 1997 11:11 | 9 | ||
Thanks John, I open a CLD. but in the meantime can the customer work like before, so not using the SKT functions, is the bug also present if dings are done this way? The customer also complains that the sockets his program opens, are not cleaned up and after a while he must shutdown and restart his NT, is this also a known problem in the final ssb kit? I found some references of a QAR (832) for a EFT version, reporting link resource problems, but do not have access to it. Jean-Paul, (Brussels) | |||||
| 223.3 | JAMIN::WASSER | John A. Wasser | Thu Apr 17 1997 11:14 | 16 | |
> in the meantime can the customer work like before, so not using the SKT > functions, is the bug also present if dings are done this way? The problem is in the DECnet stack underneath the socket library so I don't think that switching to a different socket library will help. One customer reported that they do NOT see the problem when using the WinSock 2 library but I'm not sure if that is a general work-around or not. > The customer also complains that the sockets his program opens, are not > cleaned up and after a while he must shutdown and restart his NT, is this > also a known problem in the final ssb kit? I have heard of a problem that sounds like that but I don't know if it will be fixed in the V7.0A release. | |||||