1 |
"Richard Fish" <bigfish@××××××××××.org> posted |
2 |
7573e9640609251725q52cf39efr58d0f59dfb15ce5b@××××××××××.com, excerpted |
3 |
below, on Mon, 25 Sep 2006 17:25:50 -0700: |
4 |
|
5 |
> On 9/25/06, Patric Douhane <patric@×××××.se> wrote: |
6 |
>> That didn't help. But I have some more info on my problem though, found |
7 |
>> this line when trying to emerge mozilla-firefox: |
8 |
>> |
9 |
>> "./loadmsgcat.c: 1295: internal compiler error: Segmentation fault" |
10 |
>> |
11 |
>> Hope someone can interprete it... |
12 |
> |
13 |
> Every time I've seen this message on my systems, it has been the result of |
14 |
> flakey hardware, particularly memory. Nothing stresses the memory system |
15 |
> quite like compiling. |
16 |
> |
17 |
> If the system works otherwise, I suspect flaky memory timings. Try |
18 |
> backing those off (in the BIOS), and see if the problems disappear. |
19 |
> |
20 |
> And of course, if you are overclocking anything, stop! |
21 |
|
22 |
100% agreed! I used to have some borderline generic memory, rated pc3200 |
23 |
(400 MHz), that wasn't quite stable at that. The trouble was my machine |
24 |
didn't at that time have a BIOS with memory timing limit capacities, so it |
25 |
was clocked what it was rated and that was that. |
26 |
|
27 |
After suffering with it for awhile, I discovered they had a new BIOS |
28 |
update out which allowed memory timing limits. Setting it to limit @ 183 |
29 |
MHz (DDRed to 366), I guess PC3000, it was stable as a rock, no problems |
30 |
whatsoever, until I upgraded to 8 gig awhile later. |
31 |
|
32 |
Anyway, while it would occasionally freeze the machine, doing whatever, |
33 |
the worst was bunzip2ing and compiling. Those would segfault frequently |
34 |
enough that I had to babysit all my emerges, and learn how to restart in |
35 |
the middle of them instead of starting over. So yes, definitely, if gcc |
36 |
is segfaulting, that's a very strong hint of a hardware problem. |
37 |
|
38 |
Note that memory is one possibility, but another strong candidate is bad |
39 |
power, either due to a bad (or underpowered) computer power supply, or in |
40 |
one case that came up on the lists, an underpowered UPS, or possibly |
41 |
simply bad incoming power. So in addition to checking memory, verify your |
42 |
entire power system, from the wall, thru your UPS (if you don't have one, |
43 |
try one, but make sure it's high enough powered), thru your computer power |
44 |
supply itself. Low power's the equivalent of clocking beyond stable, in |
45 |
that the effect is occasional zeros where there should be ones. And yes, |
46 |
it's gcc and bzip2 (well bunzip2) that seem most sensitive to it. |
47 |
|
48 |
On the bright side, when there was a problem, gcc would segfault or |
49 |
there'd be other errors. I never had an issue with bad builds due to the |
50 |
memory. It either built right, or it failed to complete the build at all. |
51 |
|
52 |
-- |
53 |
Duncan - List replies preferred. No HTML msgs. |
54 |
"Every nonfree program has a lord, a master -- |
55 |
and if you use the program, he is your master." Richard Stallman |
56 |
|
57 |
-- |
58 |
gentoo-amd64@g.o mailing list |