Hi John, hi Greg,
> >The current problem is with sendsize/selfcheck. If the machine tries to
> >contact itself for amcheck, I get errors in syslog about selfcheck
> >dumping core. ...
Yes, I had the same problem with Solaris 2.8.
> That should be pretty easy to track down (solid, reproducible errors are
> my favorites :-).
Same to me! ;-)
So, I ran "selfcheck" inside gdb and got a Segmentation Fault in
"client-src/client_util.c" in function "build_exclude()" here:
[...]
exclude = fopen(excl->name, "r");
--> while (!feof(exclude)) {
[...]
No big surprise -- the file handle "exclude" was used without checking
if "fopen()" succeeded (which it did not). This is fixed in the current
Amanda CVS.
If you want to continue using 2.4.3b3, apply the following patch:
*** client_util.c.orig Sun Mar 3 18:10:51 2002
--- client_util.c Mon Apr 8 20:59:14 2002
***************
*** 232,240 ****
for(excl = options->exclude_list->first; excl != NULL;
excl = excl->next) {
exclude = fopen(excl->name, "r");
! while (!feof(exclude)) {
! if(fgets(aexc, MAXPATHLEN, exclude))
! add_exclude(file_exclude, aexc, verbose);
}
fclose(exclude);
}
--- 232,242 ----
for(excl = options->exclude_list->first; excl != NULL;
excl = excl->next) {
exclude = fopen(excl->name, "r");
! if (exclude != NULL) {
! while (!feof(exclude)) {
! if(fgets(aexc, MAXPATHLEN, exclude))
! add_exclude(file_exclude, aexc, verbose);
! }
}
fclose(exclude);
}
> If that doesn't work, put this data (from the amandad*debug log file of
> the packet) in a file:
>
> OPTIONS ;
> GNUTAR /var/db/mysql 0 OPTIONS
>|;bsd-auth;compress-fast;no-record;index;exclude-list=/usr/local/lib/amanda/excl
> GNUTAR /etc 0 OPTIONS
>|;bsd-auth;compress-fast;no-record;index;exclude-list=/usr/local/lib/amanda/excl
> GNUTAR /usr/local 0 OPTIONS
>|;bsd-auth;compress-fast;no-record;index;exclude-list=/usr/local/lib/amanda/excl
>
> Then run selfcheck by hand, as the Amanda user, with this file as stdin
> and see what happens. If it fails, do the same thing under a debugger.
Thanks, John, for these instructions on running "selfcheck" alone --
using them made debugging quite straightforward. :-)
Maybe this nasty little bug in 2.4.3b3 makes a good reason for
releasing 2.4.3b4 soon? ;-)
Best regards,
Holger
--
[EMAIL PROTECTED] .................... ++49 +5246 80 1438
