Thank you for your responses. First, this is kind of "one in ten years" problem, because there should never be an I/O error on the load library PDS. The new module was copied to the production loadlib by means of our regular CCM system; this seemed to work, but the module was not usable (S106). We fixed the problem by simply sending the module to the production library again. This original problem is examined by IBM support, no solution yet.
We were astonished, too; we tried to understand, what has happened. I guess, that the entry is inserted in CDE, when the directory entry in the loadlib has been found (remember the major/minor things and the load attributes all come from the directory entry), and then, in a second step, the real "module body" is read into storage. Fortunately, the address is not put into the CDE before the read has completed successfully (for example: when getmaining the needed storage). If so, the abends would be still more crazy. Anyway, IMHO it would be better to insert the entry into the CDE after the whole module has been loaded successfully. Maybe this is not possible any more with program objects type 3. And: maybe there was some change in the past, when the program objects were introduced, which, as a side effect, changed the behaviour of LOAD in case of such errors. Another workaround for us is: don't use the address in the CDE entry, if it is zero or X'80000000'. In this case we do the LOAD again, which will lead to another S106, which is the desired behaviour. Kind regards Bernd Am Donnerstag, 26. Juni 2008 17:44 schrieben Sie: > Rick Fochtman quoted > > -------------------------<unsnip>---------------------- > > My first thought is that the invalid CDE should be removed by LOAD > > processing if the LOAD fails. But there's also a good case for fixing > > the problem that led to the S106 Abend as well. > > [<CLC>] I would be fairly astonished to find an (invalid) CDE on the JPQ > after a failed load and even more astonished to find an LLE on the > loading task. However, chasing those chains is not an official > programming interface, so IBM may be reluctant to "fix" it. > > I am guessing that if you used the official interfaces (LOAD, CSVQUERY > etc) then you would get the right answer, even in the presence of a load > failure. Peter Relson is the contents supervision guru and I suppose he > will comment if he feels like it. My guess is "working as designed". > > CC > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
