1 |
On 2013-04-22, David Relson <relson@×××××××××××××.com> wrote: |
2 |
> On Sun, 21 Apr 2013 00:44:46 +0400 |
3 |
> the guard wrote: |
4 |
> |
5 |
>> |
6 |
>> |
7 |
>> |
8 |
>> Суббота, 20 апреля 2013, 19:56 UTC от Grant Edwards |
9 |
>> <grant.b.edwards@×××××.com>: |
10 |
>> > On 2013-04-20, the guard <the.guard@××××.ru> wrote: |
11 |
>> > |
12 |
>> > > The package i decided to install required a gcc rebuild so I |
13 |
>> > > started rebuilding it and got a bus error. I've googled and found |
14 |
>> > > suggestions to lower makeopts, but it didn't help. |
15 |
>> > |
16 |
>> > Every time I've gotten bus errors when building things it turned out |
17 |
>> > to be a hardware problem. |
18 |
>> > |
19 |
>> > Bad RAM, failing CPU, failing motherboard power supply capacitors, |
20 |
>> > bad disk controller card (obviously, that was a _long_ time ago). |
21 |
>> > |
22 |
>> > If I were you, I'd start by running memtest86+ overnight. |
23 |
>> |
24 |
>> |
25 |
>> memtest revealed nothing |
26 |
> |
27 |
> We had an old QNX machine start giving bus errors during compilation of |
28 |
> a large application. Running memtest (for approx 40 hrs) showed |
29 |
> nothing, but a close visual examination of the motherboard showed |
30 |
> bulging capacitors, i.e. failing capacitors. |
31 |
|
32 |
|
33 |
"Bad caps"? Those can really give all the kinds of problems, and look |
34 |
really random. |
35 |
|
36 |
I've also seen occasions where a certain northbridge was less tolerant |
37 |
regarding voltages and would render the whole system unstable with a |
38 |
specific brand of memories (the memories were OK, but the system would |
39 |
still become unstable). |
40 |
|
41 |
There was also a more serious case where I started getting random |
42 |
segfaults with a computer, as I started leaving it on for longer and |
43 |
compiling larger programs. Apparently, the memory modules were seated in |
44 |
a less than optimal configuration, leading the motherboard to believe |
45 |
there was *another* memory module. Thing is, for several months the |
46 |
system was OK, because apparently it never needed more than the first |
47 |
half of the memory, or if it did, it did not try to use the result of |
48 |
addressing the second half. That was a lot of luck, I guess. (The less |
49 |
lucky part are the emerge -e systems anf emerge -e worlds which |
50 |
followed.) |
51 |
|
52 |
|
53 |
-- |
54 |
Nuno Silva (aka njsg) |
55 |
http://njsg.sdf-eu.org/ |