Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 30 Sep 1998 11:52:24 -0700
From:      "George W. Dinolt" <George.W.Dinolt@lmco.com>
To:        Alexander Sanda <entropy@compufit.at>
Cc:        current@FreeBSD.ORG
Subject:   Re: Very strange thing happened today
Message-ID:  <36127DE8.77225B65@lmco.com>
References:  <19980930135418.A15529@compufit.at>

next in thread | previous in thread | raw e-mail | index | archive | help
Alexander Sanda wrote:

> Ok, here is the deal:
>
> While making buildworld today (CVS'upped ca. 11am GMT), I got an error:
>
> /opt/src/src/gnu/lib/libobjc/../../../contrib/gcc/objc/list.h:142:
> parse error before `objc_list'
>
> Well, I checked the file list.h and found the following:
>
> (starts at line 139)
>
> /* Free list (backwards recursive) */
>
> static void
> list_frestruct objc_list* list)
> {
>   if(list)
>     {
>       list_free(list->tail);
>       free(list);
>     }
> }
>
> Note that the function definition is wrong in some way :)  I checked
> against the original file on ftp.freebsd.org and found out that the
> real line has to be:
>
> static void
> list_free(struct objc_list* list)
>
> Some_thing_ has stolen 2 characters from the function definition. Best
> of all, neither the filesize nor the timestamp have been changed
> (still 3521 bytes, 18 Sep 1996), _but_ 2 junk characters (0-bytes) have
> been added to the end of the file. I still have a copy of the altered
> file, so if somebody would be interested, I could mail him the header
> file.
>
> Ok, I'am a bit worried, what happened here ?
> Black magic, voodoo, aliens, cosmic rays ? :)
>
> Yes, this whole stuff sounds a bit unrealistic, but it's not a joke, it
> happened exactly this way.
>
> The problem is - of course I expected this - not reproduceable,
> and will probably never happen again, but I'am still a bit scared about
> it. I'am quite closely following -current (aka -BETA), and I'am
> building world at least once a week. However, I'am still a.out.
> Normally, I'am doing make -j16 (mainly, because it's a good stress test
> :) ) and - apart from "real" build failures due to wrong code - I never
> got strange or non-reproduceable build errors.
>
> I don't believe in bad hardware, because this box is running solid like
> a rock - I never get strange crashes, segfaults or something else, and
> I never saw a NT bluescreen on this box. But it could be a harddisk
> problem, right ?
>
> Hardware description (might be unimportant):
>
> Gigabyte GA6x86 Socket 8 mainboard, FX chipset, Pentium Pro/200 (512K),
> 192 MB EDORAM (36bit Kingston SIMMs _with_ parity)
>
> /usr/src and /usr/obj are both on the same IDE drive (Fujitsu
> MPA3035ATU).
>
> --
> # /AS/                               http://privat.schlund.de/entropy/ #
> #                                                                      #
> # XX has detected, that your mouse cursor has changed position. Please #
> # restart XX, so it can be updated.            -- From The Gimp manual #
>
> To Unsubscribe: send mail to majordomo@FreeBSD.org
> with "unsubscribe freebsd-current" in the body of the message

I have seen very similar behavior in the last day. I am running "current"
using ctm patches through src-cur.3558.
Note that the "kernel"  is running at that level. The utilities are about 3
days earlier since I hadn't tried a complete rebuild for the last 3 days,
until last night. I am running ELF, with soft updates enabled on scsi
drives (200MHz Intel, no MMX, 2940 Fast and Wide controller,  and 128M of
memory.) I have seen  three cases of corruption of the kind you reported.
In one case, the file was truncated off the front. In the second case, the
file lost two characters in the middle and transposed two other
characters.  In the final case,  the last part of the file was replaced
with data from a completely different file. In all cases, file sizes and
modification dates showed no change.

I don't have any good data to report on where the problem might be. The
problems have been sporadic. They have all occurred while I was trying a
make world (no -j). They have occurred after significant file activity as
indicated from the output of the "make world". I am reporting this only to
provide an other example  what you have seen.

George W. Dinolt



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?36127DE8.77225B65>