For a few months now, Slaven Rezić has been asking me to look at the way the Find A Tester tool keeps bring back an old email address of his, rather than the current one. The problem is the way that Metabase now stores the tester reference for a report. Instead of a specific email address it stores a resource fact, that just happens to be a tester. The tester resource fact then has a number of email addresses associated with it, except it has no way of knowing a tester's preference for which email address is most suitable. As such the list I retrieve, is essentially random. There isn't a good way to distinguish between new and old address.

In my workings on the CPAN Admin site (more news on that very soon), I spotted a gap in the knowledge a tester can tell us about themselves. Their current contact address. In many cases, particularly if they are also a CAPN author, this could be their PAUSE email address (AUTHOR@cpan.org), but equally could be a central email address they prefer to be contacted on. This then enables the tester to keep their contact information current, even if they change jobs. The mechanism behind the scenes now uses this, if available, as the primary addresss to display in the Find A Tester tool, but also now in the "raw" reports on the Reports website, and in the Reports Mailer tool that sends out the summary and report mails to authors.

There is still more to be changed to ensure that the contact address is considered the primary address, but hopefully it will get easier and more reliable to contact testers.

To find out how to update your contact address, please see tomorrow's post :)

Over the weekend of the QA Hackathon, which by all accounts was another great success (and I think David, Ricardo aand I all wish we could have been there too), various work on CPAN Testers 2.0 was underway. With the submission to the Metabase and the feed from the Metabase into the cpanstats database being two very distinct parts of the eco-system, it has made work on both parts a lot easier as it only requires one interface between them. While much of the interface has been evolving, it has for the most part been fairly straightforward. However, unfortunately while testing the interface, I noticed that several reports had been posted with no text report. Seeing as this is the part that authors really need to see, it means that we now have to discard those reports. In addition, David also found a problem with how S3 names objects, which was causing a problem for storing facts within the Metabase.

With these problems surfacing, the decision was made to restart the Metabase with a completely new instance. This is now Beta2 :) The previously submitted reports are not lost, but will be evaluated and valid reports will be inserted into the new instance once we have a little more time.

The Reports website has now been upgraded, both to the latest version of Labyrinth (with more performance improvements), and with full CPAN Testers 2.0 support. As we will shortly be moving to the Metabase as the primary feed for reports, in order to ensure that the Metabase is fully supported, all the links on the website will now start to reference the GUID rather than the cpanstats id or the old NNTP id. What this means is that all the YAML, RSS and JSON files all now use the GUID as a primary key. The 'id' field that currently exists will remain and contains the cpanstats id. If you have tools that use the 'id' you should not need to make any changes, but you may wish to think about using the GUID for display or link purposes in the future. The YAML, RSS and JSON files that currently exist are being updated as reports come in and the respective pages are rebuilt.

The Daily Summary, and associated reports that are mailed out, are currently being updated to reference the GUID. This should be rolled out later this week. After that everything should be in place to roll out the Metabase feed once the Beta testing is complete. All being well this should be really soon now ;)

Several months ago, Dave Cantrell noticed he wasn't receiving the Daily Summaries for some of his distributions that had received FAIL reports, and were shown on his Reports page. This began a series of code fixes and debugging that never seemed to resolve the issue.

With the recent changes to the CPAN Testers Statistics site, to reduce SQL calls and generally improve processing, I began the same with the Daily Summary code. Having added some additional logging information and reorganising the SQL queries, the problem Dave highlighted became clear. One of the SQL queries wasn't returning exactly what I expected, at least not all the time. In most cases it would, but you had to catch it just right to see what was happening. As a result the bug is now fixed, and Daily Summaries are flowing properly again.

The recent clean up has also highlighted another problem, that is outside of the scope of anything CPAN Testers specific, and resides with authors themselves. It amazes me how many authors give false or invalid addresses, or change their address and forget to update their PAUSE account. Is it any wonder that some have trouble trying to contact authors. At some point, using the CPAN Testers Preferences system, I plan to start publishing a list of PAUSE IDs that fail to respond. Those with full boxes won't be included, but will specifically highlight IDs that return invalid address error messages.

I see this as being a useful resource to remind authors that their PAUSE account may need updating, and also for anyone wishing to contact an author, whether there is any likelihood of them actually receiving a useful response. There are currently 7744 PAUSE accounts (at the time of writing), and although the number returning errors is quite small, it's enough to be annoying for those trying to contact those particular authors.

At the beginning of the last month, my automated mailer for authors and testers went live. For the most part many will never see the mails it produces, which is good, as it will mean that authors are uploading packages nicely formatted, and testers aren't sending out mangled reports. There have been an average of 3 or 4 authors a month who upload distributions without an archive extension accepted by CPAN/CPANPLUS or missing a version number. With testers it varies depending whether a bogus upload has filtered through the system, but again there are at least a few a month. This check now runs daily, as opposed to when I ran it manually roughly once a month. As a consequence it will hopefully help authors more quickly to know when they have uploaded a package that wasn't quite packaged correctly, and help testers spot when their test queue has got a bad package. If you do receive a mail from the system and don't believe your package/report should have been included, please let me know. It should be noted that authors can upload any archive format they choose, or any file format for that matter. However, to be most useful to end users, particularly those using the automated tools such as CPAN.pm and CPANPLUS, it is recommended that the following archive formats are used: .tar.gz, tgz, .tar.bz2 and .zip. In addition the format for a reliable upload is <code>-.</code>.

Also at the beginning of the month, we had a slight hiccup with the update to the Reports site. Unfortunately Léon was in the process of moving his server and the file used by Parse-BACKPAN-Packages became unavailable. Leon quickly sent me a copy of the file and I pointed the code at a local copy of the file instead of the remote one. However, it did prompt me to think about possibly running a mirror BACKPAN server, and thus creating the file myself. At least that way if anything happens to Leon's server, there is an alternative location from which the file can be grabbed from. Once I've got it working to my satisfaction, I'll shall implement a patch to Parse-BACKPAN-Packages so that it knows about alternative locations. As a consequence, this new BACKPAN mirror is now available publicly via http://backpan.cpantesters.org.

The CPAN Testers Wiki got an update this month too. Following some XHTML fixes and a little bit of a code clean up, I also added syndication feeds for the Recent Changes, as requested by David Golden. The fixes to the XHTML templates, CSS and RSS have hopefully helped to provide a cleaner interface.

Between 13th and 19th October there were some signification changes to the infrastructure of CPAN Testers. The fact that nobody appears to have noticed was a good thing. Expect an official announce soon :)

Over the course of the month, JJ came up with an alternative method of displaying the reports information on the Reports website. Many authors have complained (loudly), that due to the large number of reports their distributions have amassed, it gets difficult to see the important data from the irrelavent (to them). As a first attempt to minimise this the default selections within the Javascript ensured that only CPAN available distributions, tested against unpatched versions of Perl, were seen on each page. However, this also has the problem that for large listings the Javascript can still take a large amount of time to render the page correctly. For some visitors it can also be a bit disconcerting when the page initially locks while it runs the javascript, then half of the page suddenly disappears. This was always a short term measure, and following discussions with JJ, he has come up with a solution that uses a JSON data file, which loads in the browser with the predefined defaults. The new default will be to ONLY render the current distribution, with the ability to "Expand All" should you wish to see the other distribution versions. This will then mean no disappearing report data and no long running javascript. My thanks to JJ for persisting with this and figuring out the best way to render everything. The changes aren't live yet, but expect an update at some point this month.

Other website developments have continued at quite a pace and while some work has mostly been to improve XHTML compatibility and fixes cosmetic bugs, the new sites for Report Administration and Author Reporting Preferences are coming along very well. More news on those in the coming weeks. The CPAN Testers git repository has now been live for several weeks and will be expanding once more of the backend tools and code are cleaned up and packaged.

Again we had 126 testers submitting reports last month, our second highest by month, with 16 new addresses mapped, of which we had 6 new testers identified.

Work continues on for the websites and mailing systems along with numerous other changes within the CPAN Testers ecosystem. More news soon.

Well a lot has been happening this month! After a quiet revamp of the Statistics website, the work to complete the revamped Reports site continued on. At least for a few days. Following some rather large threads on the perl-qa mailing list, it was decided that part of the change for CPAN Testers, to bring about a centralised author notification system, should be brought forward. With the increased volume of reports and the fact that some authors are uninterested in the reports as a whole, a daily summary report was generated, as part of the Authors Notification System, and a new version of Test-Reporter was released, which would stop smoke testers sending reports directly to the authors. However, this turn of events also received an equal if not bigger resistance, as a number of authors requested to have their direct mails turned back on! Unfortunately, we won't be turning them back on (yet!), but we do have plans to enable the direct mails to those authors who want them, by way of the Preferences system, which will be launched as soon as we have the pieces in place. This was to be part of a major CPAN Testers 2.0 launch, but it looks like we may end up releasing parts in stages now. However, if an author wishes to still receive these emails in the short term, they can download David Golden's App-CPAN-Testers-Remailer, and configure it to read the NNTP server as and when it updates.

The new Reports site has now finally been launched, and initial feedback was positive. There are still some improvements to be made, but after 3 months of solid work, please don't expect anything soon. The biggest drawback with the current static site, is that it creates 3.7GB worth of files, and takes several hours just to generate those files that need updating each day. The dynamic site will improve on this considerably, by only caching files that are actually requested. In addition the preferences system can then use AJAX calls to update the page rather than trying to load everything up front. This will then mean that the update system can be made much more responsive, and once the full HTTP submission system is working, the websites can be updated every few hours. Though perhaps more importantly, those authors who wish to receive reports the moment they hit the server, can then open the floodgates to their own mailservers as much as they want ;)

Chris and David have been hard at work releasing updates to CPANPLUS-YACSmoke and CPAN-Reporter respectively, with David also patching CPAN.pm and CPANPLUS to implement some of the fixes to the toolchain for testers. The testing tools have moved on leaps and bounds within the last year, and setting up an automated smoke tester is now much easier than it once was. With David and Chris being much more responsive to the updates of the smoke test tools at the moment, it should be noted that CPAN-YACSmoke hasn't had much attention from Rob or myself. There is a version 0.04 in existence privately, which may see a final release, but largely the plan is to move any useful functionality into CPANPLUS-YACSmoke, where Chris can look after it more attentively.

One of the changes implemented by David, was following a clarification by Graham Barr over the use of the UNKNOWN grade. For the last few years it has been taken to mean that the test suite could not be found for the distribution, and therefore it is unknown whether the distribution passes or not. However, Graham has since stated that the grade was to also include those instances in the build stage where something goes wrong that is unclear whether the fault is with the distribution, environment, a prerequisite or something else. Previously the build stage errors were reported as FAIL, and have been a great cause of debate for authors when FAILs appear that were out of their control. The reassignment to UNKNOWN will hopefully allow authors to ignore these reports, and for testers to investigate further whether they need to fix their own environment or send a specific bug report. Following on from this, David Golden has written a "nagbot" that will be emailing testers to alert them when their testing environment need upgrading.

Another quiet launch was the CPAN Testers Development Site. This is a central resource that is updated daily with the data files and source code snapshots for the collection of CPAN Testers websites, so you have a central source to refer to. However, for the source code repositories for the whole CPAN Testers ecosystem, you should refer to the Contributing page of the CPAN Testers Wiki. Here you will find all the known repositories for the smoke testing tools and website generation tools. Patches, as always, are welcome, so if you feel a burning desire to fix some part of the system, now you know where to find the latest code versions.

Again we had over 100 testers submitting reports last month, with 23 new addresses mapped, of which we had 15 new testers identified. Rick Myers, a recent addition to the CPAN Testers community, caught out Chris with an extra 481 reports to become top tester of the month, and we've seen a slight rise in testers last month too, with 120 posting reports.

I expect a quiet a month for October, as we let the dust settle and allow David to follow Chris (nothing new there then ;)) and become a new Dad. That's not to say we will be idle, as there is still plenty to be getting on with, as we move closer to CT2.0.