angel@umigw.MIAMI.EDU (angel li) (12/05/89)
On a DECstation 3100 running Ultrix 3.1, I can't seem to issue a kill command to a running process and get a core file. The process gets the signal but a call to abort() in its signal catching routine does nothing. Has anyone seen this behavior? -- Angel Li University of Miami/RSMAS Internet: angel@flipper.miami.edu UUCP: ncar!umigw!angel
maj@cl.cam.ac.uk (Martyn Johnson) (12/06/89)
VAX Ultrix 3.0 had at least two kernel bugs which messed up core dumps. One of these provoked exactly the effect you describe; it was a vulgar bug in the access checking routines which were confused by the EINTR lying around in u.u_error. Digital fixed at least one, possibly both, of the bugs on the VAX in maintenance release 3.1 (the reason I'm not sure is that I actually fixed them myself in the source and still run 3.0). The fixes don't seem to have reached RISC Ultrix. The problem you describe is present in release 3.1A for the 5400. The fixes are trivial at the source level, so all it takes is for somebody to persuade Digital to actually release a patch or something. Martyn Johnson maj@cl.cam.ac.uk University of Cambridge Computer Lab New Museums Site Pembroke Steet Cambridge CB2 3QG UK