(j3.2006) (SC22WG5.4036) [ukfortran] LOCK/UNLOCK question

[N.M. Maclaren]

On Jun 25 2009, Malcolm Cohen wrote:
>Bill Long wrote:
>
>> What I do object to is Nick's frequent declarations that other 
>> programmers are hopeless and incompetent.
>
>I am truly surprised that you read his statements about certain things 
>being extremely difficult - so difficult that even the experts have 
>trouble with them - as being about other programmers' competency at all.

I, too, am truly surprised at that - especially as Van said the same
thing, and I pointed out that Hoare (one of THE big names in this area)
had come to the same conclusion.  Have you read Hoare's rationale for
inventing BSP?

The reasons that I said that it is error-prone is that it is too easy to
write code that has a slightly more complicated temporal structure than
you can analyse in your head, and there are essentially NO tools for
detecting errors and analysing failures.  A few will help with deadlock,
but none will help much with complicated livelock, and none will help at
all with the common mistake of assuming that A is ordered before B
because of dependencies, when there is an unexpected path that isn't
ordered.

In my experience, most exports discover such errors only when porting
their 'working' programs to a new system that has radically different
properties than the ones they have used before.  So programs that use
system-specific communication code or are used on only a few systems
often have such bugs lurking in the undergrowth.


Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email:  nmm1 at cam.ac.uk
Tel.:  +44 1223 334761    Fax:  +44 1223 334679