1 |
On Wed, 2008-07-02 at 11:13 -0500, Greg wrote: |
2 |
> Michael Haubenwallner wrote: |
3 |
> |
4 |
> >On Wed, 2008-07-02 at 08:07 -0500, Greg wrote: |
5 |
> > |
6 |
> > |
7 |
> >> |
8 |
> >> |
9 |
> >>Well, the native build does go faster and it does compile everything |
10 |
> >>pretty smoothly. However, I'm back to having troubles with Python. I |
11 |
> >>guess I'll post more specific information this time. Here's where |
12 |
> >>Python appears to be locking up. |
13 |
> >> |
14 |
> >> |
15 |
> >>This is the entry in the process table as of now. |
16 |
> >> |
17 |
> >>greg 319512 348378 62 10:50:51 pts/0 120:07 /usr/ccs/bin/as -u -mPPC |
18 |
> >>-o |
19 |
> >>build/temp.aix-5.2-2.5/usr/local/prefix-launcher/buildroot/python/Python-2.5.2/Modules/_ctypes/libffi/src/powerpc/aix.o |
20 |
> >>/tmp//ccJ3fThk.s |
21 |
|
22 |
Can you have (or give me) a look at the content of |
23 |
that /tmp//ccJ3fThk.s ? |
24 |
|
25 |
> >> |
26 |
> >>The C value is 62, so it is busy, but busy with what? It has been going |
27 |
> >>120 minutes. |
28 |
> >> |
29 |
> >> |
30 |
> > |
31 |
> >Huh? |
32 |
> > |
33 |
> > |
34 |
> I was searching the web for what would cause a system to run slower than |
35 |
> normal. On one website, they said one of the things to check was the |
36 |
> 'C' value, which is listed after the PID and Parent PID as 62. Since |
37 |
> that is a large value, it means that that process use using a lot of |
38 |
> system resources. I don't know what it's doing with them because |
39 |
> nothing's happening. (for 2 hours) |
40 |
|
41 |
Do you know 'truss' ? |
42 |
It is to trace system (libc?) calls for any (running) program: |
43 |
$ truss -p <pid> |
44 |
|
45 |
According to lslpp, it is part of 'bos.sysmgt.serv_aid' AIX package: |
46 |
$ lslpp -w /usr/bin/truss |
47 |
File Fileset Type |
48 |
---------------------------------------------------------------------------- |
49 |
/usr/bin/truss bos.sysmgt.serv_aid File |
50 |
|
51 |
|
52 |
> By the way, is '-mPPC' a correct flag? |
53 |
|
54 |
Basically yes, although some gcc (3.3.6) here uses '-mppc': |
55 |
|
56 |
$ truss -a -f -t execve gcc -c xx.c |
57 |
200400: execve("/usr/local/bin/gcc", 0x2FF224A4, 0x2FF224B4) argc: 3 |
58 |
200400: argv: gcc -c xx.c |
59 |
169120: execve("/usr/local/lib/gcc-lib/powerpc-ibm-aix5.3.0.0/3.3.6/cc1", 0x2000AC98, 0x200068C8) argc: 13 |
60 |
169120: argv: /usr/local/lib/gcc-lib/powerpc-ibm-aix5.3.0.0/3.3.6/cc1 -quiet |
61 |
169120: -D__GNUC__=3 -D__GNUC_MINOR__=3 -D__GNUC_PATCHLEVEL__=6 xx.c -quiet |
62 |
169120: -dumpbase xx.c -auxbase xx -o /tmp//cckikAyc.s |
63 |
169122: execve("/usr/ccs/bin/as", 0x2000AC98, 0x200068C8) argc: 6 |
64 |
169122: argv: /usr/ccs/bin/as -u -mppc -o xx.o /tmp//cckikAyc.s |
65 |
|
66 |
> Is this on a 64bit machine? |
67 |
|
68 |
AFAICT, the machine I work on is a 64bit hardware running a 32bit |
69 |
kernel. But I'm pretty sure we have used 64bit kernel already without |
70 |
any problems, even if we built python there. |
71 |
|
72 |
> I assume you're not running 5.2. I should |
73 |
> probably get 5.3. 5.2 has problems. |
74 |
|
75 |
I'm not aware of any problem on 5.2 (it's not that long that we have |
76 |
5.3). |
77 |
Its more the other way round: AIX 5.3 has at least two bugs I know of in |
78 |
its header files, causing global symbol constructors not being called |
79 |
with g++ when included. |
80 |
|
81 |
/haubi/ |
82 |
-- |
83 |
Michael Haubenwallner |
84 |
Gentoo on a different level |
85 |
|
86 |
-- |
87 |
gentoo-alt@l.g.o mailing list |