Profile cover photo
Profile photo
Geoffrey Sneddon
Geoffrey's posts

Just trying out GRAMPS. Trying to merge a few places, I've had three outcomes so far:
 * merging, but leaving a blank place that throws an uncaught exception if opened;
 * throwing an uncaught exception while merging; and
 * just hanging.

As tempted as I am to try and fix this, the fact that fairly obvious functionality is this diabolically broken makes me dubious about the entire quality of the product as a whole. Do they have any tests? Looking on their wiki they seem to have some unit tests and some manual tests… and it appears there's no regular running of the tests. Many units, which I'd guess are significant amounts of code, have single digit numbers of tests.

It's sad that testing remains an afterthought for so many projects, and it's something we (as a community trying to build a platform on free software) badly need to tackle, as quality issues like the above are a massive turn-off for anyone wanting to use the software.

So heartbreaking catching up on Twitter, reading +Eric Meyer  essentially tweeting the death of his daughter, a few hours beyond her sixth birthday. To have this happen to someone I quite literally grew up listening to — and whom I have countless mutual friends with — is almost unfathomable to me; the only thing more unfathomable to me is how he and his family must feel at this time of incredible loss. To Rebecca!

Doubting the use of a university education once more.

Pretty unconvinced I've actually learnt much while I've been at uni, at least insofar as CS goes. Sure, I get a piece of paper (sorry, parchment) saying I know CS, but is it really worth that? Most of what I've learnt in the past few years has been reading (mostly papers) in my spare time. I don't see how university has actually helped me much, at all…

Someone needs to remind me that reading VM papers isn't conductive to writing linguistics essays. Because I seem to be failing to remember this right now, deep in papers about interpreting SSA form code.

Post has attachment

Given I have far too many compiler guys on here, I'm trying to find anything decent about implementing a moving GC, esp. in a language like C++ where pointers abound. Halp?

Today's challenge: implement [[Call]] as a JS function (i.e., function call(func, thisArgs, args)).

 - You can rely on built-ins having their initial value when the function is defined, but cannot rely on them whatsoever when the function is called (ergo func.apply is a no-go).
 - The function must have no side-effects if func has no side-effects (ergo you cannot assign func.apply = originalApply).
 - You must not rely on any behaviour except that which is specified in ES5.

Note that this is trivial if you can clone/wrap Function.prototype.apply within the above rules.

Also: I make no guarantee as to whether this is even possible.

Post has shared content
To celebrate 300 million Opera users we share a bit of insight into what you can expect from Opera at this years Mobile World Congress!

Post has shared content
In just over a year and a half, Google's almost completely turned around its design process, and started creating beautiful apps. Here's the inside story on the new Google way.

Redesigning Google: how Larry Page engineered a beautiful revolution

To throw down a few thought's about Aaron Swartz's untimely death:

 - One cannot blame any one thing in particular for his suicide; it is rare for depression to be caused by a single reason.
 - One must question whether it is fair to judge people based upon knowledge they did not have at the time they had when they made the decision.
 - The judicial system in the US seems flawed in that there is so much weight on everything before the trial (which often ends up pretty much agreeing how the trial will play out, by the virtue of plea bargains), and so easily manipulated by money (through choice of lawyer having such a large affect). (For the curious: I maintain the view that being able to negotiate sentencing is ridiculous, as sentencing should not even be a matter of discussion pre-trial; the only concern pre-trial should be whether or not a prosecution is likely to be obtained on the evidence available (i.e., whether or not the case should be brought before a court of whether it would merely be a waste of court time as the case was not viable; it is important that the "victim" isn't involved in this choice as it avoids the victim being blackmailed into dropping the case), and once in trial the only thing to be determined should be the guilt of the accused, and only once their guilt has been established should sentencing be discussed, but quite possibly my Scottish upbringing is just showing through here).
 - That (successfully) prosecuting people is a method of career advancement is by-and-large nonsense. It gives them an inherent interest in getting people found guilty, regardless of their actual guilt.
 - Prison sentences seem astronomical in the US: that someone can get as harshly punished for copyright violations as murder seems backwards. But this seems part of a larger belief in the US that people must "pay" for their crimes through sordid conditions (at the tax-payers expense!) rather than having a sentence that compensates the copyright owners for what they have lost (in the UK, it is exceptionally rare to see trials for piracy, because the amounts that are likely to be awarded don't justify the legal expense).
Wait while more posts are being loaded