If you want to have this newsletter mailed to you or you want to make comments/suggestions about the format/content then send an email to acedb@sanger.ac.uk.
A very short newsletter for August, I was tempted to include some silly season stuff but while writing this terrible things were happening in the States.
Truly grim, sympathies and condolences to all readers in the US.
The Sanger Centre has had to tighten up security in recent months and this has prompted some changes to ftp protocols. This change has affected a number of users who regularly ftp to Sanger, ULCC users will now be affected by these changes. The following mail from one of our Systems Support staff says it all:
> We have just noticed that all the virtual hosts on the remote FTP server > at ULCC allow connections from remote hosts without a reverse DNS entry. > For security reasons this is not ideal and should be enabled. Currently > the following virtual hosts are hosted by our remote FTP server: > > ftp.sanger.ac.uk > ftp.ensembl.org > ftp.ebi.ac.uk > > We propose that we enable reverse DNS entry checks as soon as possible. > Therefore, we will change the FTP access configuration file for Sanger and > Ensembl (not EBI) at midday tomorrow (Thursday 13th). This will obviously > effect a few of our collaborators and possible lead to a few days/weeks > worth of email complaints. Sorry, for this we apologies. > > When a host does not have a valid reverse DNS entry the following error > message will be displayed: > > -- start -- > Your host does not have a reverse DNS entry. > > Connections are not permitted unless there is a valid reverse DNS. > > Please contact your network administrator to resolve this problem. > -- end --
If you work at ULCC or have queries about the changes to ftp at Sanger please could you contact Systems Support direct by sending mail to ssg@sanger.ac.uk.
A common practice in the Sanger Centre is to update the main databases via overnight queues of cron jobs which in essence use tace to parse in the database modifications. One problem with this is that all the jobs tend to be run using the same userid, so all changes by all jobs will be labelled with that same userid. This makes it difficult to establish exactly which job caused which changes.
A new command line option to tace allows the changes to be registered using an alternative label to that of the current userid. This means that each cron job could be given its own label so that changes made by that cron job can be identified.
Use tace like this:
tace -tsuser new_label /your/database
You can also use this option on the command line for xace as well.
The right button menu that appears when you click on a dna or peptide homology now includes an option to efetch the sequence for that homology.
Sometimes (hard to believe I know), acedb crashes because of some fundamental error either in the code (or even in the hardware) that is detected by the operating system. When this happens the operating system does two things: it sends the program a special "signal" that the error has happened, the program can ignore this or perhaps do some cleaning up before exitting, if the program takes no action then the operating system will then try to take a copy of the program in memory from which its possible to see what the program was doing when it crashed. The copy is held in the familiar "core" file.
By default acedb intercepts the "signal" from the operating system so it can close the database properly before asking the operating system to make a core file. The core file produced by this is often useless for debugging however as its initial entries are about the section of code to close the database, and it can be impossible to get back from there to the original point of failure.
More useful is to turn off this signal catching by acedb so that the operating system will produce a core file that contains exactly the point of failure, To do this you need to:
An attempted optimisation in AQL introduced a bug which has now been fixed.
(This article is courtesy of John Sturdy jcgs@sanger.ac.uk)
Consider the following fragment of aql:
select paper, author1, institute1, author2, institute2
from paper in class Paper,
author1 in paper->Author,
institute1 in author1->Institute,
author2 in paper->Author,
institute2 in author2->Institute
where (institute1 = "MRC-LMB") and (institute2 = "CNRS")
AQL used to use a spurious attempt at optimisation which, in this example, resulted in institute1 being substituted for institute2, which resulted in no objects matching the query (since institute1 could not be equal to two different things simultaneously). This bug is now fixed.
In some cases, this change may cause unwanted large cross products. If this is a problem please send bug reports to jcgs@sanger.ac.uk.
This function is hardly ever used, which is good because it could not have worked correctly: it handled neither toggle or colour editors, but instead called graphTextScrollEntry() on them both producing garbage in whatever window they were displayed in. The function now correctly updates the state of the displayed editor boxes.
The message window did not handle multiline messages correctly. Multiline messages are now displayed on a single line with a <TAB> added in place of each newline character. This leaves the messages as readable and avoids the window becoming full with large multiline messages.
The "Signal Catching" item (from the "Admin.." button on the main window) got out of synch with the actual state of signal catching in xace. This is now fixed.
bug fixes to WWWdisplay() and addition to look for Web_location in the class Class object if not found in the object itself
Several small bugs were fixed in peptide translation:
Last months xace did not display simple dna objects in an fmap because of a trivial error in the checkSequenceLength() function which detects when users ask to view very big sequences and gives them the chance to cancel the fmap display.
Recent changes to Blixem included a removable toolbar, this introduced a but whereby the window could not be collapsed when the toolbar was "docked", this has now been fixed.
You can pick up the monthly builds from: