On Tue, 5 Jun 2007, Wolfgang Pfeiffer wrote:
Eep, can you redo that with strace -f (to follow the child)?
Yes. I've attached the output.
Ah, much better...
3943 open("/usr/lib/libnss_db.so.2", O_RDONLY) = 4
3943 read(4,
"\177ELF\1\2\1\0\0\0\0\0\0\0\0\0\0\3\0\24\0\0\0\1\0\0\23"...,
512) = 512
3943 fstat64(4, {st_mode=S_IFREG|0644, st_size=23000, ...}) = 0
3943 mmap(0xf6f9000, 88012, PROT_READ|PROT_EXEC,
MAP_PRIVATE|MAP_DENYWRITE,
4, 0) = 0xf6f9000
3943 mprotect(0xf6ff000, 61440, PROT_NONE) = 0
3943 mmap(0xf70e000, 4096, PROT_READ|PROT_WRITE|PROT_EXEC,
MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 4, 0x5000) = 0xf70e000
3943 close(4) = 0
3943 access("/etc/ld.so.nohwcap", F_OK) = -1 ENOENT (No such file or
directory)
3943 open("/usr/lib/libdb-4.3.so", O_RDONLY) = 4
3943 read(4,
"\177ELF\1\2\1\0\0\0\0\0\0\0\0\0\0\3\0\24\0\0\0\1\0\1\274"...,
512) = 512
3943 fstat64(4, {st_mode=S_IFREG|0644, st_size=1022832, ...}) = 0
3943 mmap(0xf5dc000, 1099776, PROT_READ|PROT_EXEC,
MAP_PRIVATE|MAP_DENYWRITE, 4, 0) = 0xf5dc000
3943 mprotect(0xf6d2000, 65536, PROT_NONE) = 0
3943 mmap(0xf6e2000, 16384, PROT_READ|PROT_WRITE|PROT_EXEC,
MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 4, 0xf6000) = 0xf6e2000
3943 mmap(0xf6e6000, 10240, PROT_READ|PROT_WRITE|PROT_EXEC,
MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0xf6e6000
3943 close(4) = 0
3943 munmap(0x30030000, 65926) = 0
3943 --- SIGSEGV (Segmentation fault) @ 0 (0) ---
I'm guessing you have compat for passwd/shadow in /etc/nsswitch.conf
If so, can you replace compat with files (or do you actually make use of
NIS/NIS+ groups)?
The real problem here is that libdb4.5 on ppc is downlevel - it has no
ELF symbol versioning. And since sendmail was built with libdb4.5, it
will cause the problem you see when libnss-db later loads db4.3 :(
Unfortunately, the powerpc build daemon is current not quite upto snuff,
and things are lagging.
I've got someone building a new libdb4.5, and then I might be able to
get sendmail built on my box.
This'll likely be a few days, however - so do try changing compat to
files if you can.
Not yet, the trace will hopefully help alot... though, iirc, you are
running a custom built kernel; can you possibly try with a stock Debian
kernel (they come in 32/64 bit UNI/SMP, and even a Prep).
If nothings else helps: Yes. I don't like the stock Debian kernels.
Yeah, I tend to not use them myself - Though I always have the latest
one 'in case of emergency' :)
However, I just got the ppc box up, so haven't as yet built my own.
But I booted the machine with various self-built other kernel images:
All versions being around 2.6.17/2.6.18, IIRC: I always got an error
similar to the one already reported, i.e. "Segmentation fault", IIRC,
of sendmail when trying to get up while booting.
Right, it is a library issue
Is there a special config option needed to be enabled in recent
kernels for this new sendmail version? If yes I'd prefer to compile my
own kernel, with the needed switch on - I intend to do this anyway
since quite some time, since 2.6.18 is really getting old now, perhaps
2.6.21 is in unstable (I'm running it now).
PS:
I just removed the second authinfo line that let sendmail complain
about it, did a
# makemap hash /etc/mail/authinfo < /etc/mail/authinfo
# sendmailconfig
and the latter ended with a second crash, like that:
Yes, the two issues are completely unrelated.
--
Rick Nelson
* athener calls Amnesty International House of Pancakes
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
