home *** CD-ROM | disk | FTP | other *** search
- Submitted-by: salevin@dal.mobil.com (S. A. Levin [Stewart])
-
- I noticed a few months ago that every once in a while my pipelined jobs
- on our RS6000 would break when I tried to put them into the background
- under csh. A few days ago IBM customer support tracked down what was
- happening. They say that under POSIX if a signal (e.g. SIGTSTP) is
- received during a read() system call that has not yet received any
- characters, the read must return an -1 count and set EINTR. (This is
- not specific to csh.)
-
- To me this is an unacceptable interaction between job control and POSIX
- compliance. In particular, the possibility of losing hours of work (as
- I did once!) is not acceptable. What would it take to solve this
- problem?
-
- Stew Levin
- Mobil
- salevin@dal.mobil.com
-
- Volume-Number: Volume 30, Number 75
-
-