Re: ANN: refdb-0.9.9-pre3 available for testing

Hi all,

this is to summarize the OS X related problems that I've been working
on lately with the kind help of Jeremy.

Jeremy Malcolm writes:
 > Something is askew again on Mac OS X.  Like with 0.9.8,[0] I just get 
 > "error" from runbib after a long pause. Previously with this problem, I 
 > would at least get a log from the back end to trawl through.  But now I 
 > don't even get that.  "refdbd -s -e 0 -l 7" produces no output after 
 > "server waiting n_max_fd=5".  So I guess the lack of debugging is the 
 > biggest problem here.  The underlying problem may be more simple.
 > 

While I wasn't able to reproduce the lack of a log file, it was pretty
apparent that the bibliography processing time exceeded the
client-side timeout (default: 600s) by a couple of minutes, resulting
in the "error" message. I have to add to my defense that these OS X
tests ran on fairly old hardware. A similar setup on my FreeBSD box
created the same 800+ reference bibliography in less than three
minutes, not anywhere close to a timeout.

Anyway, there were two reasons why the processing time increased
substantially from 0.9.9-pre2 to pre3. First, the database schemas of
both refdb and the reference database now have a bunch of new columns
to retrieve, which slows down the operation a bit. Second, the
bibliography style migration instructions were incomplete and, in
part, just plain wrong. Jeremy's attempt to migrate his custom style
caused most in-text citations to use the author "Anonymous", which in
turn led to the problem that all these references had to be
distinguished by a year suffix as in (Anonymous, 2002ah). The unique
suffix is selected by a recursive function which uses one SQL query
per iteration. As you may guess, having 50-odd references by
"Anonymous" in the same year increases processing times quite a bit.

As a consequence, the style migration instructions in the file
"UPGRADING" were corrected and enhanced, and there is also some more
information in the documentation which ships with each source
tarball. If you use custom styles, please do read these instructions
to avoid major hassles.

While working on an OS X box I came across another PPC specific
bug. The style export was entirely broken. This has also been fixed in
the subversion sources.

 > Also, I have a feature request: although the includearg variable can be 
 > set in the runbib script, it would be better if it could be set in a 
 > configuration file.  That way, it doesn't get lost whenever the user 
 > upgrades refdb.  (I need it set in order to raise jade's maximum error 
 > count since I have many references and the normal "reference to 
 > non-existent ID" bootstrapping errors come up for each of them.)
 > 

I've addressed this problem by adding configuration files for runbib
and for refdbjade. Among other tweaks you can now increase or disable
jade's maximum error limit.

All these changes will be available in the upcoming 0.9.9-pre4 which I
hope to announce shortly.

regards,
Markus

-- 
Markus Hoenicka
[email protected]
(Spam-protected email: replace the quadrupeds with "mhoenicka")
http://www.mhoenicka.de


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/