Ken Moffat wrote:
> On Tue, Jul 24, 2007 at 10:56:00PM -0400, Joe Ciccone wrote:
>
>> The builds I've done so far have worked pretty smoothly, before I
>> committed the change I successfully finished a alpha, x86, x86_64, and
>> x86_64-64 build. I didn't do any in depth testing of the tools, but I
>> installed openssh & wget on each one and could netboot each one.
>>
>
> Well, that's plenty of testing - good to know it works for you.
>
>
>> I'll
>> rebuild my x86_64-64 and see if I can break it. If the segfaults seem
>> random, data is probably being corrupt somewhere. Did you check the disk?
>>
>>
>
> I've just fscked it - all fine. There was nothing in my syslog,
> but I've just started a long test run of smartctl.
>
>
Well, I managed to recreate this and I almost have this figured out...
Sort of. I went through all the dependencies of the commands that were
seg faulting and found that bash was the offending package. If bash was
compiled natively Glibc would work, but cross-compiled the make would
seg fault. Looked into the config.cache differences for a native and a
cross-compiled build. Here were the differences worth looking at.
(Original file was the broken version).
-ac_cv_func_mmap_fixed_mapped=${ac_cv_func_mmap_fixed_mapped=no}
+ac_cv_func_mmap_fixed_mapped=${ac_cv_func_mmap_fixed_mapped=yes}
-ac_cv_func_strcoll_works=${ac_cv_func_strcoll_works=no}
+ac_cv_func_strcoll_works=${ac_cv_func_strcoll_works=yes}
-ac_cv_func_working_mktime=${ac_cv_func_working_mktime=no}
+ac_cv_func_working_mktime=${ac_cv_func_working_mktime=yes}
-bash_cv_func_sigsetjmp=${bash_cv_func_sigsetjmp=missing}
+bash_cv_func_sigsetjmp=${bash_cv_func_sigsetjmp=present}
-bash_cv_getcwd_malloc=${bash_cv_getcwd_malloc=no}
+bash_cv_getcwd_malloc=${bash_cv_getcwd_malloc=yes}
-bash_cv_job_control_missing=${bash_cv_job_control_missing=missing}
+bash_cv_job_control_missing=${bash_cv_job_control_missing=present}
-bash_cv_printf_a_format=${bash_cv_printf_a_format=no}
+bash_cv_printf_a_format=${bash_cv_printf_a_format=yes}
-bash_cv_sys_named_pipes=${bash_cv_sys_named_pipes=yes}
+bash_cv_sys_named_pipes=${bash_cv_sys_named_pipes=present}
-bash_cv_ulimit_maxfds=${bash_cv_ulimit_maxfds=no}
-bash_cv_under_sys_siglist=${bash_cv_under_sys_siglist=no}
-bash_cv_unusable_rtsigs=${bash_cv_unusable_rtsigs=yes}
+bash_cv_ulimit_maxfds=${bash_cv_ulimit_maxfds=yes}
+bash_cv_under_sys_siglist=${bash_cv_under_sys_siglist=yes}
+bash_cv_unusable_rtsigs=${bash_cv_unusable_rtsigs=no}
-gt_cv_int_divbyzero_sigfpe=${gt_cv_int_divbyzero_sigfpe='guessing no'}
+gt_cv_int_divbyzero_sigfpe=${gt_cv_int_divbyzero_sigfpe=yes}
So, I put this in config.cache and now it works.
ac_cv_func_mmap_fixed_mapped=yes
ac_cv_func_strcoll_works=yes
ac_cv_func_working_mktime=yes
bash_cv_func_sigsetjmp=present
bash_cv_getcwd_malloc=yes
bash_cv_job_control_missing=present
bash_cv_printf_a_format=yes
bash_cv_sys_named_pipes=present
bash_cv_ulimit_maxfds=yes
bash_cv_under_sys_siglist=yes
bash_cv_unusable_rtsigs=no
gt_cv_int_divbyzero_sigfpe=yes
I'd still like to find out which one of those is the offending entry.
But for now, it's atleast narrowed down quite a bit.
_______________________________________________
Clfs-dev mailing list
[email protected]
http://lists.cross-lfs.org/cgi-bin/mailman/listinfo/clfs-dev
