headers
Paul A. Hoadley | 1 Apr 2005 04:37
Favicon
Gravatar

Re: Upgrade bootstrap problem

Hi Markus,

On Thu, Mar 31, 2005 at 10:43:05AM +0200, Markus Hoenicka wrote:

> I've added the meta information to the refdb database only recently
> (in an amazing lack of foresight :-( ). With any 0.9.6 prerelease
> you can do a "select * from t_meta" to see all information that you
> would like to see.

Nice feature.  :-)

> 1) Do you see any differences in the SQL dumps of refdb and refdb1?

There were diffs, but nothing helpful (to me).

> 2) If I recall correctly, I've never changed the refdb schema. All I
> ever did is add new columns, tables, or data. This should allow to
> create a refdb database with any recent version of RefDB and install
> the old dump file over that database. This will give a boatload of
> errors (duplicate data, tables and columns already exist etc) but it
> should add any missing information. You might try that as you have
> backup data anyway.

I was going to try this, but a better solution turned up by chance.  I
found an old, broken copy of the XML stylesheet.  I also tracked down
a mail from you that contained a patch to correct the errors.  :-)  I
applied the patch, loaded the stylesheet, and everything works again.

Thanks for your reply, not to mention the patch.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Jorgen Olsen | 1 Apr 2005 10:01
Picon

SV: refdbd does not respond to the getref command

> I login to the machine from a Putty terminal under SSH.
>

Do you have a chance to work directly on the Linux box? Again, it shouldn't
matter, but sometimes strange things happen.

> The problem comes when I try to fetch a complete reference:
>
> refdbc: getref -d jopub ':AU:=olsen'
> aborting...
>
> Actually the terminal hangs and I have to issue a ctrl-C command to abort.
>

Well, if you see the "aborting..." message, the terminal should not hang. It
is
the client waiting for data from the server. Of course nothing should hang
at
all, but I suspect it is rather the server than the client that hangs.

> I then inspected the tables at the Mysql server. The refdb and the jopub
> databases are created and the only thing that looked strange was that the
> fields for title, abstract and a few others were of the type BLOB. I
changed
> this to the type TEXT but the problem persists.
>

You should leave the columns as BLOB. MySQL has the peculiar feature that
TEXT
fields are case-preserving, but not case-sensitive. This sometimes alters
(Continue reading)

Permalink | Reply | 3 comments |
headers
Markus Hoenicka | 1 Apr 2005 10:48
Picon

Re: SV: refdbd does not respond to the getref command

Hi,

Jorgen Olsen <jolsen <at> imbg.ku.dk> was heard to say:

>
> I include here the refdbd.log file below. The session covers the successful
> getau command and the failed getref command. I hope you can diagnose
> something out of it !
>
> 6:pid=32257:Fri Apr 01 07:57:14 2005:adding client 127.0.0.1 on fd 6
> 6:pid=32257:Fri Apr 01 07:57:14 2005:server waiting n_max_fd=6
> 6:pid=2371:Fri Apr 01 07:57:14 2005:serving client on fd 6 with protocol
> version
> 6:pid=2371:Fri Apr 01 07:57:14 2005:dbi is up
> 6:pid=2371:Fri Apr 01 07:57:15 2005:child finished client on fd 6
> 6:pid=32257:Fri Apr 01 07:57:14 2005:parent removing client on fd 6
> 6:pid=32257:Fri Apr 01 07:57:14 2005:server waiting n_max_fd=5
> 6:pid=32257:Fri Apr 01 07:57:15 2005:child exited with code 0
> 6:pid=32257:Fri Apr 01 07:57:15 2005:server waiting n_max_fd=5
> 6:pid=32257:Fri Apr 01 07:57:27 2005:adding client 127.0.0.1 on fd 6
> 6:pid=32257:Fri Apr 01 07:57:27 2005:server waiting n_max_fd=6
>

If you indeed start refdbd as you stated (refdbd -s -e 0 -l 7) you should get a
far more verbose output. Could you please try again? Please kill all instances
of refdbd (as seen by "ps ax|grep refdbd") and restart it with the above
command. I take from your log that the child process which handles the getref
command does not finish. It apparently does not crash (you would see a "child
exited with code 0" as well), but it does not do you any good either.
(Continue reading)

Permalink | Reply | 2 comments |
headers
Jorgen Olsen | 1 Apr 2005 11:09
Picon

SV: SV: refdbd does not respond to the getref command


Hi,

Markus

There were actually two refdbd processes that I were not aware of. After
killing them I get a much more useful output (for you I hope): (I have X'ed
out the password)

refdbd -s -e 0 -l 7
dbi_driver_dir went to:

dbi is up using default driver dir
Available libdbi database drivers:
mysql
application server started
use /tmp/refdbd_fifo2650 as fifo
server waiting n_max_fd=4
adding client 127.0.0.1 on fd 5
server waiting n_max_fd=5
try to read from client
parent removing client on fd 5
server waiting n_max_fd=4
serving client on fd 5 with protocol version 2
201-28-62-46
send pseudo-random string to client
getau "olsen" -u jolsen -w 033047106116082061048078 -d jopub
dbi is up
localhost
jolsen
(Continue reading)

Permalink | Reply | 1 comment |
headers
Diwaker Gupta | 1 Apr 2005 12:25
Picon
Gravatar

Re: xnote woes

Hi list, Markus

On Nov 5, 2004 3:08 PM, Markus Hoenicka <markus.hoenicka <at> mhoenicka.de> wrote:
> Diwaker Gupta writes:
>  > 7:pid=874:Thu Nov 04 09:29:41 2004:UPDATE t_note SET
>  > note_content='<pre>\n            This is a test.\n            L
>  > ets see how it does.\n        </pre>' WHERE note_id=1
> 
> Think I'm getting old. This is a bug in the sqlite driver. I've fixed
> this bug a couple of weeks after the latest libdbi-drivers release and
> forgot about it. That's why it works for me but fails for everyone
> else.
> 
> To fix this problem, please get the CVS version of libdbi-drivers,
> build and install it. The escaped newlines should then be gone.

Sorry for digging old graves, could you tell me if this has been fixed
in the latest release of libdbi? Right now I have version 0.7.1-2
(Ubuntu), but the problem still seems to be there. Do I absolute
*need* to checkout the CVS?

--

-- 
Diwaker Gupta
http://resolute.ucsd.edu/diwaker

-------------------------------------------------------
This SF.net email is sponsored by Demarc:
A global provider of Threat Management Solutions.
Download our HomeAdmin security software for free today!
http://www.demarc.com/info/Sentarus/hamr30
(Continue reading)

Permalink | Reply | 2 comments |
headers
Michael Smith | 1 Apr 2005 12:49

Re: Debianising RefDB

Markus Hoenicka <markus.hoenicka <at> mhoenicka.de> writes:

[...]

> I think it is possible to provide Debian packages for consumption by apt-get
> even before the packages are an official part of Debian. On the client side all
> you need is an appropriate entry in your apt config file. I'll try to find out
> what I have to provide on the server side in order to make this work.

I would be glad to help with that if you want. Please let me know.
I could set it up for you at http://refdb.sourceforge.net/debian/
(or whatever other subdirectory), and it wouldn't take long.

I have a Debianized repository at http://docbook.sf.net/debian/
now (at /home/groups/d/do/docbook/htdocs/debian/ from the shell),
so feel free to take a look at that.

Basically, all you need are the following files in the same
directory as your *.deb package(s):

  Release
  Packages.gz

If you want to make source versions of the package apt-gettable,
you also need a Sources.gz file and the *.diff.gz, *.dsc, and
*.orig.tar.gz that form the source "package".

The Packages.gz and Sources.gz files are generated using the
Debian dpkg-scanpackages and dpkg-scansources commands. It is best
to generate them on the server side, but it's also possible (if
(Continue reading)

Permalink | Reply | 13 comments |
headers
Markus Hoenicka | 1 Apr 2005 13:08
Picon

Re: SV: SV: refdbd does not respond to the getref command

Hi,

Jorgen Olsen <jolsen <at> imbg.ku.dk> was heard to say:

> SELECT DISTINCT t_xauthor.refdb_id FROM t_xauthor, t_author WHERE
> t_author.author_id=t_xauthor.author_id AND t_xauthor.xauthor_type='primary'
> AND t_author.author_name = 'olsen'

You should be aware that this query is unlikely to return a match. You test for
a name 'olsen' which will be matched literally. You probably want to use
"getref :AU:~Olsen" or "getref :AU:=Olsen,J." instead. However, this query
should return 0 matches instead of doing strange things.

> free(): invalid pointer 0x5896c0!
> free(): invalid pointer 0x5896e0!
> free(): invalid pointer 0x589700!
> free(): invalid pointer 0x589720!
> child exited with code 0
> server waiting n_max_fd=4
>
>
> I guess that it is the "invalid pointer" message that is the key to the
> problem - does it make any sense to you ?
>

These messages indeed are the problem, although I'll have to check precisely
where they come from. It is still possible that the code is not 64bit-safe for
some reason. I do not have any experience yet with the implications of a 32bit
to 64bit migration. I'll see what I can do over the weekend.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Markus Hoenicka | 1 Apr 2005 13:11
Picon

Re: xnote woes

Diwaker Gupta <diwakergupta <at> gmail.com> was heard to say:

> Sorry for digging old graves, could you tell me if this has been fixed
> in the latest release of libdbi? Right now I have version 0.7.1-2
> (Ubuntu), but the problem still seems to be there. Do I absolute
> *need* to checkout the CVS?
>

Unfortunately yes. There has been no new libdbi release since then. I'm
currently working on a couple of minor fixes for MySQL > 4.01 which will
require a new release anyway. Lets hope that I'll be able to motivate the other
libdbi developers to crank out a new release.

regards,
Markus

--

-- 
Markus Hoenicka
markus.hoenicka <at> cats.de
(Spam-protected email: replace the quadrupeds with "mhoenicka")
http://www.mhoenicka.de

-------------------------------------------------------
This SF.net email is sponsored by Demarc:
A global provider of Threat Management Solutions.
Download our HomeAdmin security software for free today!
http://www.demarc.com/info/Sentarus/hamr30
Permalink | Reply | 1 comment |
headers
David Nebauer | 1 Apr 2005 15:57
Picon

Re: Debianising RefDB

David Nebauer writes:

> > 6.  The /usr/share/refdb/db/refdb database
> > 
> > Currently the post-install script looks for the default refdb database.  
> > If not found, it is created from the sqlite dump file.
> > 
> > I think it would be good to include a post-install check of any existing 
> > refdb database.  This check would ensure the database is compatible with 
> > the new version of refdb (during an upgrade, for example).  If not 
> > compatible, the user is warned.

It has only just occurred to me to wonder exactly _what_ to tell a user 
who has just upgraded refdb to a version with a new db format.

The traditional approach (with a source install) is to instruct the user to:
1. Dump their reference data, styles etc.,
2. Install new refdb,
3. Delete old refdb system database,
4. Create new refdb system database, and
5. Import old reference data, styles, etc.

This is no longer possible -- after the deb install the damage already 
is done.  It is too late to backup their data for re-import since refdbd 
will no longer run.

Anybody got any ideas?  This can't be the first time this problem has 
occurred with a debian package.  Surely database apps like 
mysql|postgresql etc. have coped with format changes over the years.
(Continue reading)

Permalink | Reply | 1 comment |
headers
David Nebauer | 1 Apr 2005 14:28
Picon

Re: Debianising RefDB

Hi Michael,

Michael Smith wrote:

>Markus Hoenicka <markus.hoenicka <at> mhoenicka.de> writes:
>
>>I think it is possible to provide Debian packages for consumption by apt-get
>>even before the packages are an official part of Debian. On the client side all
>>you need is an appropriate entry in your apt config file. I'll try to find out
>>what I have to provide on the server side in order to make this work.
>>
>
>I would be glad to help with that if you want. Please let me know.
>
I've offered to have a go at setting this up and will certainly take you 
up on your offer of help if I get stuck.

Initially the repository will be debs only (no source debs) and i386 only.

The tools seem simple to use.  All I have to do now is decipher the 
directory structure...

>Anyway, let me know if you'd like for me to help with this.
>

Will do, and thanks.

Regards,
David.
(Continue reading)

Permalink | Reply | 12 comments |

Gmane