Re: portsdb -Uu results in coredump

2004-09-06 Thread Brandon S. Allbery KF8NH
On Tue, 2004-09-07 at 01:08, Michael Nottebrock wrote:
> bdb_hash, then run portsdb -fu - works for me, so that also points to be some 
> bug in bdb1.

Hrm.  I thought bdb1 btrees were well known to be buggy, to be honest. 
(I had a feeling it was going to turn out to be this issue as soon as
someone reported that configuring portupgrade to use bdb_hash fixed it.)

-- 
brandon s. allbery[linux,solaris,freebsd,perl] [EMAIL PROTECTED]
system administrator  [WAY too many hats][EMAIL PROTECTED]
electrical and computer engineering, carnegie mellon univ. KF8NH

___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


Re: portsdb -Uu results in coredump

2004-09-06 Thread Michael Nottebrock
On Tuesday 07 September 2004 05:13, Garance A Drosihn wrote:

> I think the problem is really in bdb
> (which ruby is calling for database-work), and not in ruby itself.

You can actually change the database backend manually: Install 
databases/ruby-bdb and set the envvar PORTS_DBDRIVER to bdb_btree or 
bdb_hash, then run portsdb -fu - works for me, so that also points to be some 
bug in bdb1.

-- 
   ,_,   | Michael Nottebrock   | [EMAIL PROTECTED]
 (/^ ^\) | FreeBSD - The Power to Serve | http://www.freebsd.org
   \u/   | K Desktop Environment on FreeBSD | http://freebsd.kde.org


pgpARu9iZOhcU.pgp
Description: PGP signature


Re: portsdb -Uu results in coredump

2004-09-06 Thread Garance A Drosihn
At 10:49 PM -0400 9/6/04, Sahil Tandon wrote:
kstewart wrote:
There is a bug in ruby that shows up as a bus error. Follow the
topic on [EMAIL PROTECTED] There are several ways to alter the INDEX[-5]
so that it occurs less frequently.
Occurs less frequently?  That's not what I'm looking for.  Is there
a *fix* for the root cause?
Oddly enough, there are a lot of other people who would also prefer
a more complete fix.  If we had a fix, we would install it and you
wouldn't have to work around the problem.  There are developers who
are looking into the problem.  I think the problem is really in bdb
(which ruby is calling for database-work), and not in ruby itself.
--
Garance Alistair Drosehn=   [EMAIL PROTECTED]
Senior Systems Programmer   or  [EMAIL PROTECTED]
Rensselaer Polytechnic Instituteor  [EMAIL PROTECTED]
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


Re: portsdb -Uu results in coredump

2004-09-06 Thread kstewart
On Monday 06 September 2004 07:49 pm, Sahil Tandon wrote:
> kstewart wrote:
> > There is a bug in ruby that shows up as a bus error. Follow the topic on
> > [EMAIL PROTECTED] There are several ways to alter the INDEX[-5] so that it occurs
> > less frequently.
>
> Occurs less frequently?  That's not what I'm looking for.  Is there a
> *fix* for the root cause?

I haven't seen one. I have seen comments but when ruby is updated, you will 
probably be seeing a fix.

>
> > I started using portindexdb because it doesn't error off. I also don't
> > use categories, which it doesn't produce.
>
> I'm using portsindexsb as a work around, but it still doesn't address or
> solve the problem which makes portsdb -Uu fail.  I've tried removing and
> re-extracting the ports tree and doing whatever else was suggested in
> the -ports@ mailing list(s) - to no avail.
>

I am also running portindexdb for that reason. For what I use INDEX-5.db for, 
portindexdb works just fine. We may have to wait until the port freeze is 
over. So far, I am only having to use in on my RELENG_5 machine. It cvsups 
port-all 2x a day as a cronjob and moving a "#" from one command ot another 
in the script isn't much work :).

Kent

-- 
Kent Stewart
Richland, WA

http://users.owt.com/kstewart/index.html
Support the Bison at http://www.buffalofieldcampaign.org/
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


Re: portsdb -Uu results in coredump

2004-09-06 Thread Sahil Tandon
kstewart wrote:
There is a bug in ruby that shows up as a bus error. Follow the topic on 
[EMAIL PROTECTED] There are several ways to alter the INDEX[-5] so that it occurs less 
frequently. 
Occurs less frequently?  That's not what I'm looking for.  Is there a 
*fix* for the root cause?

I started using portindexdb because it doesn't error off. I also don't use 
categories, which it doesn't produce.
I'm using portsindexsb as a work around, but it still doesn't address or 
solve the problem which makes portsdb -Uu fail.  I've tried removing and 
re-extracting the ports tree and doing whatever else was suggested in 
the -ports@ mailing list(s) - to no avail.
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"



Re: portsdb -Uu results in coredump

2004-09-06 Thread Günther Dippe
I had to download a fresh ports.tar.gz (and extract it).
Then portsdb -Uu works.
Cheers
Günther
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


Re: portsdb -Uu results in coredump

2004-09-05 Thread kstewart
On Sunday 05 September 2004 11:12 pm, Maxim Maximov wrote:
> Daniel O'Connor wrote:
> > On Mon, 6 Sep 2004 15:14, Maxim Maximov wrote:
> >>That doesn't help. I tried to entirely delete all /usr/ports, ruby*,
> >>portupgrade*, removed /var/db/pkg/pkgdb.db, install those all over again
> >>and still get a coredump.
> >>
> >>SIGSEGV appears in libc.so.4 in __bt_split() and it seems like a real
> >>bug in there triggered by ruby_bdb1 and some line in the ports/INDEX.
>
> Oops, I meant SIGBUS here.
>
> > Hmm, I am running 6-current (libc.so.5) so perhaps the patch to fix it
> > wasn't tested in 4.x?
>
> What patch are you talking about? I've got coredumps on two my 4.10
> machines and all my 5.x/6 systems didn't get this error, BTW.

There is a bug in ruby that shows up as a bus error. Follow the topic on 
[EMAIL PROTECTED] There are several ways to alter the INDEX[-5] so that it occurs less 
frequently. 

I started using portindexdb because it doesn't error off. I also don't use 
categories, which it doesn't produce.

BTW, if you follow ports-all, you are also supposed to also follow ports@ 
where this failure has been discussed for several days now. It may seem like 
a waste of time until something like this pops up.

Kent

-- 
Kent Stewart
Richland, WA

http://users.owt.com/kstewart/index.html
Support the Bison at http://www.buffalofieldcampaign.org/
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


Re: portsdb -Uu results in coredump

2004-09-05 Thread Maxim Maximov
Daniel O'Connor wrote:
On Mon, 6 Sep 2004 15:14, Maxim Maximov wrote:
That doesn't help. I tried to entirely delete all /usr/ports, ruby*,
portupgrade*, removed /var/db/pkg/pkgdb.db, install those all over again
and still get a coredump.
SIGSEGV appears in libc.so.4 in __bt_split() and it seems like a real
bug in there triggered by ruby_bdb1 and some line in the ports/INDEX.
Oops, I meant SIGBUS here.

Hmm, I am running 6-current (libc.so.5) so perhaps the patch to fix it wasn't 
tested in 4.x?

What patch are you talking about? I've got coredumps on two my 4.10 
machines and all my 5.x/6 systems didn't get this error, BTW.

--
Maxim Maximov
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


Re: portsdb -Uu results in coredump

2004-09-05 Thread Martin Hudec
On Mon, Sep 06, 2004 at 01:45:18AM -0400 or thereabouts, Tim Kellers wrote:
> I'm trying it again, though this time it's on Yet Another 5.3gbeta2 machine.  
> I'll try it again on a production 4.10-Stable machine in a few minutes.
> 
> Tim
> 

I was receiving such errors from portsdb -uU too (abort core dumped), 
and I switched over to portindex utility (portindex && portindexdb, 
then portupgrade -airR to upgrade my ports). Now I am using it to my
full satisfaction.

Portindex utility is located in:
/usr/ports/sysutils/portindex

Advantage of it is, that it is much much faster than portsdb.

Disadvantage of it is, that it needs python and posgresql
(at least client installed, full server is not required).


Cheers,

Martin Hudec

-- 
Martin Hudec| corwin at aeternal.net
| corwin at web.markiza.sk
http://www.aeternal.net | cell +421 907 303 393



pgpIQLwXyXsV0.pgp
Description: PGP signature


Re: portsdb -Uu results in coredump

2004-09-05 Thread Daniel O'Connor
On Mon, 6 Sep 2004 15:14, Maxim Maximov wrote:
> That doesn't help. I tried to entirely delete all /usr/ports, ruby*,
> portupgrade*, removed /var/db/pkg/pkgdb.db, install those all over again
> and still get a coredump.
>
> SIGSEGV appears in libc.so.4 in __bt_split() and it seems like a real
> bug in there triggered by ruby_bdb1 and some line in the ports/INDEX.

Hmm, I am running 6-current (libc.so.5) so perhaps the patch to fix it wasn't 
tested in 4.x?

-- 
Daniel O'Connor software and network engineer
for Genesis Software - http://www.gsoft.com.au
"The nice thing about standards is that there
are so many of them to choose from."
  -- Andrew Tanenbaum
GPG Fingerprint - 5596 B766 97C0 0E94 4347 295E E593 DC20 7B3F CE8C


pgpGh80OQQiv4.pgp
Description: PGP signature


Re: portsdb -Uu results in coredump

2004-09-05 Thread Tim Kellers
I'm trying it again, though this time it's on Yet Another 5.3gbeta2 machine.  
I'll try it again on a production 4.10-Stable machine in a few minutes.

Tim

On Monday 06 September 2004 01:11 am, Sahil Tandon wrote:
> Tim Kellers wrote:
> > I pkg_delete'd portupgrade and all of ruby, reinstalled portupgrade (and
> > all of ruby) and had to do a portsdb -Uu even after I did a make
> > fetchindex in /usr/ports to make the pkgdb in /var/db/pkg happy.  Since
> > I've done all of those things, portupgrade is working again.  I had to to
> > this on three machines --2 5.3 beta2 and one 4.10-STABLE, although on the
> > stable machine I managed to skip the portsdb -Uu step after the make
> > fetchindex step. Whatever broke the pkgdb did it around the 8500 number
> > port in the index.
>
> I'm having no such luck.  I've done the following:
>
> # pkg_delete -f portupgrade\*
> # pkg_delete -f ruby\*
> rmdir: /usr/local/share/ri/1.8/system: No such file or directory
> pkg_delete: unexec command for '/bin/rmdir -p
> /usr/local/share/ri/1.8/system' failed
> rmdir: /usr/local/share: Directory not empty
> pkg_delete: unexec command for '/bin/rmdir -p
> /usr/local/share/ri/1.8/site' failed
> pkg_delete: couldn't entirely delete package (perhaps the packing list is
> incorrectly specified?)
> # cd /usr/ports/sysutils/portupgrade/ ; make install clean
> # cd /usr/ports ; make fetchindex
> # portsdb -Uu
>
> ... and then I get the same ruby coredump after all that.  FWIW, I'm
> running 4.10-STABLE.  And I also encounter the error somewhere after the
>   8000th number port.
>
>
> ___
> [EMAIL PROTECTED] mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "[EMAIL PROTECTED]"
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[EMAIL PROTECTED]"