Random ramblings

Wednesday November 30, 2005

GNU libtool, Sun Studio and 64bit builds

Many free software projects, particularly ones which use autoconf/automake build files, use GNU libtool as a wrapper around compiler and linker. Unfortunately libtool doesn't pass the "-xtarget= " (e.g -xtarget=generic64), nor most other SOS10 specific link+compile flags, to ld when linking. This then will cause the build to fail if the objects were built 64bit, as the linker will try link against 32bit system libraries. A quick workaround is to set CC to 'cc -xtarget=....'. Other flags which must be passed at link stage can also be put here (-xipo, -xildoff, etc.). This apparently is fixed in libtool 1.5.20. Another workaround is to pass these flags using -Wl.

Course, all too often Free software projects will not build at all with anything but GCC, sadly.

Rossi on four wheels

Rossi beats former WRC champion Colin McRae at a rallying show, beats former karting champion at another show (Rossi's first time out in a kart with a gearbox). Course, they're just former champions, anyone could beat them really.

He's also been testing with Ferrari again, 3 days at Mugello, a tougher circuit than Ferrari's test track at Fiorina. Rossi's fastest time at the latter apparently is 2s off Schumachers lap record there, so obviously Rossi is completely uncompetitive.

Be interesting to see what happens in 2007 or 2008, whether Rossi goes or not. If he does, don't bet against him becoming F1 champion (if only because he won't jump unless he's certain he'd be competitive).

C'etait un Rendezvous

"C'etait un Rendezvous" is a 1978 short film by Claude Lelouch of a car driving through Paris, by an unknown driver. By common repute the car is a Ferrari 275 GTB¹ and reaches speeds of over 220km/h² in the film. The speed claim is endorsed in this analysis by some students.

The 220km/h claim doesn't seem right though. Go and watch the video, watch it like you would watch the road normally when driving, try and identify hazards and traffic lights as early as possible at least and think about what you would do as a driver (if you had licence to drive so recklessly). Do that before reading the spoiler in the next paragraph.

It doesn't feel that fast, does it? Even on the long stretch from l'Arc down to Place du Concord, you still have plenty of time to see traffic lights, even other traffic. On the stretch after Place du Concord to the Louvre (169s to about 200s), where the speed analysis linked to above claims the highest speed of 220km/h is reached, other traffic on the road is seen with plenty of time to spare and even the armchair observer watching the video has time to think "better change lane to avoid that car". It does not feel at all like 220km/h.

So out of curiosity, and with scepticism, I had to try to verify the 220km/h speed between 171s and 189s of the link above. Using this image from Google maps the distance between the first and second bridges after Concord, shortly after which the car turns into the Louvre, is about 430m, according to scale given (the section of road is highlighted in white). The distance between Concord and the first bridge is noticeably shorter, but call it 400m. According to the analysis above, the distance between 171s (after Concord) to 189s (well before the second bridge) is over a kilometre, 1.1km. However, according to Googles' satellite image, it's less than 800m.

The average speed, using the Google image scale, from where he turns onto the road after Concord at 169s to the first bridge at 181s is:

400/(181-169)*3.6 = 118km/h

From there to the second bridge at 195s is:

430/(195-181)*3.6 = 108km/h

So, if Googles' sense of scale is to believed (and apparently they know how to scale ), the analysis above is incorrect and, at least by this methodology, the 220km/h is disproven.

Of course, Google could be wrong and the data the students used could be correct. However, we can estimate the speed of the car by measuring the relative speed as he passes the two cars by the second bridge. The first car looks like a Renault 5, the second is a Citroen 2CV and appears to be, at most, 5 car lengths ahead of the Renault 5, so about 18m at most. The car in the film takes 3s to pass the 2CV after passing the Renault:

18/3*3.6 = 22km/h

The convenient factor here is that a 2CV struggles to get much past 100km/h. Which would, assuming the 2CV is near to flat-out, put the speed of "our" car at roughly 125km/h maximum. This figure consistent with the speed of 108km/h derived from the Google image data, and strongly suggests the 220km/h analysis must be flawed.

It is quite possible the car reached higher instantaneous speeds after Concord before it reached the Renault, say at around 181s, but that's not what the analysis linked to above tried to measure, and even then it seems highly unlikely to me those speeds were anywhere near 220km/h. You could calculate an estimation of this by measuring the rate of increase in apparent size of the Renault 5 as it first appears in the distance and some basic geometry. I suspect that too would work out to between 120km/h and 140km/h absolute tops.

220km/h though, no way!

1. The car used possibly was Claude's Mercedes, rather than a Ferrari. This picture possibly confirms it, supposedly of Claude and the gyro-stabilised camera rig on the front of a Mercedes.

2. For the metrically-challenged, 220km/h is about 137mph, 120km/h ~= 74mph, 100km/h ~= 62mph.