Gfortran does not compile dynamo (and others)

[Tim Prince]

Brooks Moses wrote:
> Ignacio Fernández Galván wrote:
>> I checked a bit more about the error. The unit that's trying to open
>> (variable "In") is number 5, and inquire says it's open before (normal)
>> with a blank filename. With the simple test I cannot reproduce the
>> error, but I was thinking maybe it was some conflict between processes.
>> Gaussian is made of different executables ("links") which are called
>> one after the other (but not by a script, the time taken by each link
>> is cumulative, it seems), could it be that some other process has
>> opened unit 5, and then there's an error when trying to open it again?
>> But it works closing it explicitly... humm.
>
> Unit 5 is standard input.  It's automatically opened.
>
> This is one of the reasons that it's very common advice on 
> comp.lang.fortran to never use unit numbers below 10 in a program.  
> Most compilers use 5 for standard input and 6 for standard output, so 
> you definitely want to avoid those, but other low numbers are fairly 
> common for this (and some also have a unit number assigned to standard 
> error, as well) and using them is likely to cause conflicts of this 
> nature on other compilers.
>
> In any case, the best solution is to use some unit number other than 
> 5, but explicitly closing unit 5 at the beginning of the program will 
> probably also work -- though closing it, on some compilers, will cause 
> problems if you try to write to "*".  (I think gfortran is immune to 
> that, but not all compilers are.)
>
This is another example of why certain compilers close an implicitly 
opened unit when it is OPENed explicitly.  Granted, OPEN(5) is not going 
to work portably among various compilers.