oester+@rchland.ibm.com (Bob Oesterlin) (10/24/90)
The background queuemail daemon gives messages like the following when doinglocal delivery in an AFS environment: <fdplumb>File '/afs/rchland.ibm.com/usr1/ent/Mailbox' (opened by open) isstill open. ( 7:20:43 PM ) <fdplumb>Total of 1 open files. ( 7:20:43 PM ) Delivery succeeds, but the messages bother me. Plumber is trying to tell mesomething! What can I do to track down the problem? This only occurs on my AIX3.1 systems, not my rt_aos4 or rt_aix221 systems. Bob
nsb@THUMPER.BELLCORE.COM (Nathaniel Borenstein) (10/24/90)
Something is very odd, conceivably a bug in the fdplumber stuff itself. The thing is, the file it says is open is a directory, yet it says it was opened by "open" rather than by "opendir". Very peculiar. One approach might be to look through all the queuemail code for any place where any "open" call might actually be issued on a directory rather than on a file. Does AIX 3.1 do anything weird with opening & closing directories? If, for example, it renamed "closedir" to something else, then the fdplumber stuff needs to be augmented to recognize the new name for it; otherwise it is simply missing the closes, and printing out spurious messages about directories that really have already been closed. -- Nathaniel
Craig_Everhart@TRANSARC.COM (10/24/90)
The message is probably coming from trymail, not queuemail. Trymail doesn't do an opendir of a Mailbox directory. However, as you recall, vmail() (a routine from overhead/mail/lib -> libmail.a) will look to see if its destination directory is in AFS; the IsOnVice() procedure takes an open file descriptor, so vmail() typically open's and close's the Mailbox directory itself around making this check. But, Bob has tinkered with his copy of vmail.c on rs_aix31 because the AFS-dependent ioctl (VIOCIGETCELL) doesn't work in the AFS port to rs/6000's yet. Until he tells me otherwise, I'll assume that Bob's tinkering eliminated the close() done in overhead/mail/lib/vmail.c, line 117 or 122. Craig
oester+@rchland.ibm.com (Bob Oesterlin) (10/29/90)
*Blush*... Craig hit the nail right on the head! I #ifdef'd out the close at line 117. Never mind.... Bob