| David Ham 2004-11-17, 3:59 pm |
| On Wed, 17 Nov 2004 16:30:29 +0000 (UTC)
helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to
reply) wrote:
> In article <cndeup$3m1$1@online.de>, helbig@astro.multiCLOTHESvax.de
> (Phillip Helbig---remove CLOTHES to reply) writes:
>
>
> Wouldn't it be a good idea for the standard to require that a compiler
>
> give an error if it is accessing modules via USE compiled with a
> previous version of the compiler? It seems to me that the only
> alternative is to recompile everything after each compiler upgrade,
> since I'm sure the standard doesn't require the results of compiling
> modules to be compatible for all versions of the compiler.
It'd probably be a good idea if compilers noticed old versions but
making the standard mandate that would require the standard to specify
huge volumes of things which it doesn't currently specify.
For starters, the processor (ie the thing processing the fortran source)
isn't required to be a compiler at all. It has been frequently pointed
out that it would be completely legal to make a Fortran interpreter.
>
> In my case, I got a strange error which forced me to track down the
> problem. But perhaps I could have gotten just a wrong answer, which
> would have been worse due to the danger of not noticing it.
>
> What do folks do? Recompile everything you USE after upgrading the
> compiler? If not, does your compiler vendor guarantee compatibility?
>
I use the Intel Compiler for Linux and we often seem to be able to go
through minor version number upgrades without ill effect but for big
upgrades we have to recomile everything.
David
|