filePro with no stdin
Fairlight
fairlite at fairlite.com
Tue Jan 24 00:40:30 PST 2006
Confusious (Bob Rasmussen) say:
> I suggestion was made earlier that I think bears repeating: try running
> some sort of a trace on the offending process. It should show you what
> I/O operation it is executing, and what value is being returned by the OS.
Okay, I borrowed a client's system to do what's been requested.
Short version: read() fails on fd 0 with EBADF (no great surprise), and
ioctl() fails on fd 0 with ENOTTY (also no great surprise). Repeatedly.
10MB worth of trace in 5 seconds (give or take a second). Thankfully it
gzip's down to 54K.
Full trace is given through the point when I sent SIGINT 5 seconds in.
Don't just take my word for it; anyone that wants the trace can take a
gander. Trace is available at:
http://members.iglou.com/fairlite/fprace.txt.gz
CC to fP-Tech support. [Note to fP-Tech support: This details the system
trace of dclerk in a race condition where file descriptors 0, 1, and 2 are
all closed (fd 0 being the one causing the race condition), as previously
reported by me.]
Trace platform used:
fP 5.0.14 on RHEL3.
Trace methodology:
strace -o fprace.txt dclerk 0>/dev/null 1>/dev/null 2>/dev/null
Bests,
mark->
More information about the Filepro-list
mailing list