*BSD News Article 22935


Return to BSD News archive

Xref: sserve comp.os.linux.development:2275 comp.os.386bsd.development:1341
Newsgroups: comp.os.linux.development,comp.os.386bsd.development,alt.os.bsdi
Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!haven.umd.edu!news.umbc.edu!europa.eng.gtefsd.com!uunet!mcsun!news.inesc.pt!dec4pt.puug.pt!matuc2.mat.uc.pt!rps
From: rps@matuc2.mat.uc.pt (Rui Pedro Mendes Salgueiro)
Subject: Re: ugly name for core dumps (core.imagename) -> patch for "img.core"
Message-ID: <1993Oct27.170050.1573@matuc2.mat.uc.pt>
Organization: Dep. Matematica Universidade de Coimbra
X-Newsreader: Tin 1.1 PL3
References: <13326@dirac.physics.purdue.edu>
Date: Wed, 27 Oct 1993 17:00:50 GMT
Lines: 23

This discussion has already taken place in the BSD/386 mailing list
(some of the participants included ex-CSRG members).
Conclusions (according to my recollection):

    1 - 'name.core' is better than 'core.name', but now is a little
	late to change things.
    2 - 'core.name.pid' is dangerous. The reason to use 'core.name'
	instead of 'core' is not to make it easier to discover which
	program has dumped core (use file(1) for that) but to prevent a
	core from some daemon to overwrite other cores. The problem
	with 'core.name.pid' is that a daemon launched from init or
	inetd that always dumped core could fill the root partition.
	And if you want to be sure that no core is overwritten the
	proper way would be something like 'core.name.time_in_us'
	(pids wrap around after ~30000).

For BSD/386 someone has made a "commitee solution" patch.
With this patch the behaviour is selectable when the kernel is configured.

--
 Rui Salgueiro |   Dpt. de Matematica    |"In my life / Why do I smile
 rps@mat.uc.pt | Universidade de Coimbra | at people who I'd much rather
 rps@inescc.pt |    Portugal - Europe    |   kick in the eye" - Morrissey