Building CUnit from Source

I haven’t had to build CUnit myself for a while, because most of the systems I work with have it in their packaged software repositories, but for Solaris it’s not there, and it turns out, it’s quite painful to build!

  • First, the latest version on sourceforge, 2.1.3, is incorrectly bundled, with files ( missing, so you have to use 2.1.2.
  • Second, the directions in the README don’t include all the details of what autotools commands need to be run.

Here’s the commands I finally used to get a build. Note that you do need to run libtoolize to get some missing support scripts installed, and that you need to also run automake in “add missing” mode to get let more support scripts. Then and only then do you get a build.

tar xvfj CUnit-2.1-2.tar.bz2
cd CUnit-2.1-2
libtoolize -f -c -i \
  && aclocal \
  && autoconf \
  && automake --gnu --add-missing \
  && ./configure --prefix=/usr/local \
  && make \
  && make install

Update: The bootstrap file does provide the required autotools flags.

The Tyranny of Environment

Most users of PostGIS are safely ensconsed in the world of Linux, and their build/deploy environments are pretty similar to the ones used by the developers, so any problems they might experience are quickly found and removed early in development.

Some users are on Windows, but they are our most numerous user base, so we at least test that platform preemptively before release and make sure it is as good as we can make it.

And then there’s the rest. We’ve had a passel of FreeBSD bugs lately, and I’ve found myself doing Solaris builds for customers, and don’t get me started on the poor buggers running AIX. One of the annoyances of trying to fix a problem for a “weird platform” user is just getting the platform setup and running in the first place.

So, having recently learned a bit about vagrant, and seeing that some of the “weird” platforms have boxes already, I thought I would whip off a couple vagrant configurations so it’s easy in the future to throw up a Solaris or FreeBSD box, or even a quick Centos box for debugging purposes.

I’ve just been setting up my Solaris Vagrantfile and using my favourite Solaris crutch: the OpenCSW software repository. But as I use it, I’m not just adding the “things I need”, I’m implicitly choosing an environment:

  • my libxml2 is from OpenCSV
  • so is my gcc, which is version 4, not version 3
  • so is my postgres

This is convenient for me, but what are the chances that it’ll be the environment used by someone on Solaris having problems? They might be compiling against libraries from /usr/sfw/bin, or using the Solaris gcc-3 package, or any number of other variants. At the end of the day, when testing on such a Solaris environment, will I be testing against a real situation, or a fantasyland of my own making?

For platforms like Ubuntu (apt) or Red Hat (yum) or FreeBSD (port) where there is One True Way to get software, the difficulties are less, but even then there is no easy way to get a “standard environment”, or to quickly replicate the combinations of versions a user might have run into that is causing problems (libjson is a current source of pain). DLL hell has never really gone away, it has just found new ways to express itself.

(I will send a psychic wedgie to anyone who says “docker”, I’m not kidding.)

What to do about Uber (BC)

Update: New York and Chicago are exploring exactly this approach, as reported in the New York Times: “Regulators in Chicago have approved a plan to create one or more applications that would allow users to hail taxis from any operators in the city, using a smartphone. In New York, a City Council member proposed a similar app on Monday that would let residents “e-hail” any of the 20,000 cabs that circulate in the city on a daily basis.”

Nick Denton has a nice little article on Kinja about Uber and how they are slowly taking over the local transportation market in cities they have been allowed to operate.

[](" imageanchor="1" style="clear: right; float: right; margin-bottom: 1em; margin-left: 1em;)

it’s increasingly clear that the fast-growing ride-hailing service is what economists would call a natural monopoly, with commensurate profitability… It’s inevitable that one ride-sharing service will dominate in each major metropolitan area. Neither passengers nor drivers want to maintain accounts with multiple services. The latest numbers on [Uber], show a business likely to bring in nearly $1bn a month by this time next year, far ahead of any competitor

BC has thus far resisted the encroachment of Uber, but that cannot last forever, and it shouldn’t: users of taxis in Vancouver aren’t getting great service, and that’s why there’s room in the market for Uber to muscle in.

Like Denton, I see Uber as a mixed bag: on the one hand, they’ve offered a streamlined experience which is qualitatively better than the old taxi service; on the other, in setting up an unregulated and exploitative market for drivers, they’ve sowed the seeds of chaos. The thing is, many of the positive aspects of Uber are easily duplicable by existing transportation providers: app-based dispatching and payment aren’t rocket science by any stretch.

As an American, Denton naturally reaches for the American solution to the natural monopoly: regulated private enterprise. In the USA, monopolists (electric utilities, for example) are allowed to extract profits, but only at a regulated rate. As Canadians, we have an additional option: the Crown corporation. Many of our natural monopolies, like electricity, are run by government-owned corporations.

Since most taxis are independently owned and operated anyways, all that a Crown taxi corporation would need to do is provide a central dispatching service, with enough ease-of-use to compete with Uber and its like. The experience of users would improve: one number to call, one app to use, no payment hassles, optimized routing, maybe even ride sharing. And the Crown corporation could use supply management to prevent a race to the bottom that would impoverish drivers and reduce safety on the roads.

There’s nothing magical about what Uber is doing, they are arbitraging a currently inefficient system, but the system can save itself, and all its positive aspects, by recognizing and reforming now. Bring on our next Crown corporation, “BC Dispatching”.

Spatial IT and the Enterprise

My latest, a webinar on how OpenGeo Suite and Spatial IT go together for enterprise geo.

My Tax Dollars at Work?

Some days I feel like I’m king of the world, and some days I feel like the weak little fool that I am, and today is one of the latter.

I have written previously about the upcoming systems re-write in the British Columbia natural resources sector. This is a big one. Over $100M will be spent on IT services and software. Before you spend that kind of money, you should be clear about what the business value is, about why this big moon-shot of a project is the right thing to do, relative to the other options.

You should prepare a business plan. I wonder what that looks like?

I know something about data management in the natural resources sector, so I was actually interested more in the implementation plans than the business plans, I FOI’ed:

Any business case documents or presentations; System development timelines or development plans; Software or product comparisons

Potentially, this could be a lot of material, at least the FOI person thought so. They suggested I narrow my request:

…they’ve stated that the first option, “starting from the original request but limiting to only documents delivered to Doug Say”, would be the best option for narrowing the request and making the fee smaller, as he would be the only one conducting the search and this would significantly reduce the amount of time that would be spent on searching for records…

So, they were OK with my more limited request, that’s great. Until they send back the fee estimate:

Task Cost
28.5 hour(s) to locate and retrieve records @ $30.00/ hour $855.00
31.5 hours(s) to produce records @ $30.00/hour $945.00
31.5 hour(s) to prepare and handle records for disclosure @ $30.00/hour $945.00
1 disc(s) @ $4.00/disc $4.00
2400 scanned electronic copies of paper records @ $0.10/page $240.00
Total $2989.00

Fortunately, they accept VISA, Mastercard and Amex. Actually, no. Chastened, I reduce my request yet further:

I feel like I’m fishing with hand grenades here. Can I just have a copy of the business plan and delivery plan (project milestones and expected deliverables?) without an overview I’ll never figure out what to rationally ask for.

So now we’re down to 1-2 documents, no need for a big expensive search, the fee dogs are called off. No doubt, I’ll get my documents shortly. The deadline rolls around:

I am writing to ask that you grant a 30 business day extension to the above-noted FOI request. The additional 30 days are required in order to obtain further information from the Ministry as to the status of the records in their submission to Cabinet.

OK, I’m not a total a**hole, I can be patient here. And heck, they are making a good faith effort so ensure I can see the docs by checking them against cabinet submissions. What an awesome crew. 30 business days go by:

We are nearing the end of the review of the records for this request, however we will require some additional time to have it completed. I am writing to request that you grant a 10 business day extension so that we may provide a complete response to this request.

Again, I’m a man of infinite patience, particularly when a “complete response” is in the offing! By all means, squire, take your 10 business days! 10 business days go by:

Please see the attached documents in response to your FOI request with the Ministry of Forests, Lands and Natural Resource Operations.

Oh, frabjous day! Calooh! Calay! Let’s open up our giftie and see what the FOI elves have brought!

  • A signature page.
  • A contacts page.
  • The first 3 lines of the table of contents, the remainder redacted because it constitutes “policy advice or recommendations”. That’s right, the table of contents is policy advice.
  • The first page of the executive summary.
  • The first page of the “narrative” section (with the middle part of that page redacted for “policy advice” reasons).
  • The first page of the “investment context” section (with the end part of that page redacted for “policy advice” reasons).

That’s it! The remaining 185 pages of this 204 page business case are withheld! You want to know the reasoning behind the expenditure of over $100M on IT services and software? Too f***ing bad, chump.

Yep, feeling totally played for the chump. They will spend that $100M, by god, and they will spend it without anyone outside the “yes, minister, yes deputy, yes director” echo chamber seeing the rationale or commenting on it.

Update: I did vent a little back at the FOI officer (which, let’s be clear, isn’t really fair, he’s just the messenger here):

I’m concerned that this “process” is a farcical waste of your time. If you’re going to redact the entire document, save yourself the effort and tell me ahead of time, “BTW, the whole thing will be redacted”. Why even pretend, at this point? Who reviewed this document and decided on the redactions?

To which he patiently replied:

I apologise for not making you aware ahead of time that very little information would be released. I was hoping by this time that more information could be provided to you, however the Ministry informed that nothing has been implemented with regard to the initiatives, so the majority of the records are required to be withheld [em. mine] at this time. Should you wish to request these records at a later time, please feel free to contact me before submitting your request to determine their status and whether or not more information would be made available at that time.

I’m not sure how the implementation or not of things bears into their releasability, perhaps an FOI expert could drop some wisdom in the comments. It seems like the ultimate Catch-22, “we’ll explain the reasoning behind our decisions once they are set in stone and cannot be changed”.