1 |
Daniel Iliev wrote: |
2 |
> Petter Haggholm wrote: |
3 |
> |
4 |
>> The subject is fairly descriptive. Often -- but not always -- an |
5 |
>> emerge will render my system unusable. At a PORTAGE_NICENESS of 3, and |
6 |
>> fairly standard MAKEOPTS of "-j2" (on a single-core system), I'm ... |
7 |
>> well, rather surprised, confused, and very frustrated. I also can't be |
8 |
>> very specific, because I have no real idea of what's causing this, and |
9 |
>> the symptoms are very general: The system becomes unresponsive, the |
10 |
>> mouse will move but with enough of a lag that physically moving it may |
11 |
>> not cause a cursor movement for the next 30 seconds or so, clicking a |
12 |
>> taskbar window may not have an effect at all; sometimes I can't even |
13 |
>> ssh into the system from my other computer (to kill the emerge) |
14 |
>> because it's slow enough that the ssh daemon times out my login |
15 |
>> attempt. This never used to happen, I'm not aware of using any weird |
16 |
>> settings, I recently rebuilt my entire system with conservative CFLAGS |
17 |
>> (-march=athlon64 -O2 -pipe -fomit-frame-pointer), and I have an x86 |
18 |
>> (pentium-m) system with very, very similar settings that exhibits no |
19 |
>> such problems. |
20 |
>> |
21 |
>> Any hints, thoughts, or suggestions would be very much welcomed. |
22 |
>> System info follows. |
23 |
>> |
24 |
>> |
25 |
> 1) Does the system go out from this state after some time or you have to |
26 |
> cold reset it? |
27 |
> 2) What happens if you put PORTAGE_NICENESS=19, MAKEOPTS of "-j2 -l1" ? |
28 |
> l5 (small "L", not the number "one"), means "loadavg=<1" If loadavg goes |
29 |
> up to 1 make waits this level to drop before continuing its job |
30 |
> 3) Is DMA enabled for your HDD(s)? (hdparm -d1 /dev/xxx)? |
31 |
> 4) Have you done any overclocking? |
32 |
|
33 |
1) It does eventually, but it can take a long time. If I have work to |
34 |
do, I generally can't afford to outwait it. If I run an overnight build, |
35 |
the system will be fine in the morning, but some apps will typically |
36 |
have died (killed by GNOME for not responding? I don't know; usually |
37 |
Thunderbird will die, and sometimes Firefox as well -- thank heavens for |
38 |
Tab Mix Plus and session crash recovery. |
39 |
2) I did not know of make's -l option, but -l1 seems like a very |
40 |
reasonable setting -- I will try that when I don't need to be working |
41 |
(hopefully later tonight) and see what it does for me. Thanks! |
42 |
3) Yep, both hard drives have DMA enabled. |
43 |
4) Not an overclocker; I value stability much higher than performance. |
44 |
|
45 |
Without having tried it yet -- I need to finish some work first -- I'm |
46 |
curious about the MAKEOPTS thing, especially -j?. The other reply (so |
47 |
far) in this thread suggests -j1 and asserts that "[-j2] has no sense |
48 |
anyway, if you have no multicore or don't use distcc...". This is in |
49 |
contrast to all the recommended settings I have seen so far -- standard |
50 |
procedure seems to be -j$(no. cores+1), I presume so that one make |
51 |
process can do some CPU work while the other is I/O bound? I suppose one |
52 |
could keep the system more responsive at the cost of longer compiles by |
53 |
running only one make process, but I have never heard of anyone else |
54 |
having problems with -j2 on a single-core system ... |
55 |
|
56 |
(In the hope that you're reading this, too, Pawel -- I don't think I |
57 |
need to spam the list with an extra reply to so short an email -- I |
58 |
don't use KDE at all, so the kdeenablefinal flag doesn't really matter |
59 |
anymore; it's a leftover from a time long gone when I used some packages |
60 |
that pulled in KDE dependencies. I do use gcc 4[.1.1[-r1]], of course, |
61 |
but it's working fine with -j2 on my [x86] server and [x86] laptop.) |
62 |
|
63 |
-- |
64 |
Petter Häggholm |
65 |
-- |
66 |
gentoo-amd64@g.o mailing list |