Title: | DECmcc user notes file. Does not replace IPMT. |
Notice: | Use IPMT for problems. Newsletter location in note 6187 |
Moderator: | TAEC::BEROUD |
Created: | Mon Aug 21 1989 |
Last Modified: | Wed Jun 04 1997 |
Last Successful Update: | Fri Jun 06 1997 |
Number of topics: | 6497 |
Total number of notes: | 27359 |
As mentioned in note 5728 and 5594 of this conference we are using the mcc_mir routines within our TBL-project very extensively to store our instance- and attribute-data in the private MIR. Creation of attribute- and instance-repositories is done with the routine mcc_mir_create_repository of DECmcc V1.3.0 (ULTRIX) with the appropriate repository types MCC_K_DB_FTYPE_ENTDEF MCC_K_DB_FTYPE_ATTRIBUTE by our own FM and AM. Doing this in parallel within two processes the following problems occur: Problem 1: start process 1 start process 2 create inst-repos 1 ==> repos_id = 10 create inst-repos 2 ==> repos_id = 11 create inst-repos 3 ==> repos_id = 12 create inst-repos 4 ==> repos_id = 12 (!!) create inst-repos 5 ==> repos_id = 13 create inst-repos 6 ==> repos_id = 13 (!!) We assume, that process 2 gets the right repos_id the first time creating one repository, but after that numbering of the repos_id is done seperately by each process, that means the same repos-ids are reused. This causes confusing and unacceptable behaviour and inconsistence of the stored data. Problem 2: create inst-repos 1 ==> repos_id = 22 create inst-repos 2 ==> repos_id = 23 create attr-repos 4 ==> repos_id = 23 (!!) Reading and writing of these repositories causes the following return status: MCC_S_WRONGREPOSTYPE 52876290 Is this bug known by anybody, is there a bug-fix? Heidi
T.R | Title | User | Personal Name | Date | Lines |
---|---|---|---|---|---|
5981.1 | Sounds like a bug. | TAEC::LAVILLAT | Mon May 09 1994 11:15 | 15 | |
Heidi, Your are not supposed to create MIR repositories very often, that is why *maybe* the access to the repository table is not well protected. I would suggest you enter a CLD for this problem, so that the MCC maintenance team is aware of it. I would also suggest that you work around the problem by doing the mutual exclusion yourself via flock or any other way. Regards. Pierre. |