Return to BSD News archive
Path: sserve!manuel.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!sun-barr!sh.wide!wnoc-tyo-news!news.u-tokyo.ac.jp!ecc-s401!g150267
From: g150267@kecc.ecc.u-tokyo.ac.jp (Tomoaki)
Newsgroups: comp.os.386bsd.bugs
Subject: Fatal signal 6
Message-ID: <1810@ecc-s401.secc.ecc.u-tokyo.ac.jp>
Date: 2 Mar 93 04:42:10 GMT
Sender: news@secc.ecc.u-tokyo.ac.jp
Organization: Educational Computer Centre,the Univ.of Tokyo,Japan.
Lines: 53
Nntp-Posting-Host: kintaro
the cc1 getting signal 6 seems a bug of sscanf()
and the conceerned routines.
the program enquire sais sscanf gets signal for
a great number.
this was copiled with gcc-2.3.3 and with a small change
bellow. becouse the set/longjmp does not saves status of
i387 .
the following is a part of the massage from enquire.
*** WARNING: sscanf caused a trap
scanning: 1.7976931348623168e+308 format: %le
I do suppose the problem is the same point when cc1 gets
fatal signal 6.
so I'm *not* running 386bsd on ibm clone ,but a port of 386bsd
to pc98. Thus there might be some difference. but think that
the problem is the same.
Tomoaki Nishiyama
g150267@kecc.ecc.u-tokyo.ac.jp
*** enquire.c Sun Feb 28 18:27:29 1993
--- enquire.c.orig Thu Feb 25 10:01:39 1993
***************
*** 11,19 ****
Copyright (c) 1988, 1989, 1990 Steven Pemberton, CWI, Amsterdam.
All rights reserved.
- Change by Tomoaki Nishiyama:
- Clear exception flags when received a signal, Tomoaki Nishiyama,Feb 93.
-
Changes by Richard Stallman:
Undef CHAR_BIT, etc., if defined in stdio.h, Richard Stallman, Aug 90.
In EPROP, avoid a <= old if bad is set, Richard Stallman, May 91.
--- 11,16 ----
***************
*** 481,490 ****
jmp_buf lab;
Procedure overflow(sig) int sig; { /* what to do on over/underflow */
signal(sig, overflow);
- #ifdef i386
- asm("fclex");
- #endif /*if we don't clear the exeption flag, an exeption will reported
- in time of the next floating point instruction */
longjmp(lab, 1);
}
--- 478,483 ----