[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

970210: Promula

Darin,

>Date: Mon, 10 Feb 1997 11:17:47 -0500 
>From: Darin Ernst <address@hidden>
>Organization: PPPL(?)
>To: address@hidden
>Subject: Promula 
>Keywords: 199702101524.IAA27231 

In the above message you wrote:

> Dear Steve,
> 
> About a year ago, you helped me work out some problems with the
> netCDF using Promula under Linux.   I have been using netCDF
> for some time.   Recently, I noticed on the UNIDATA web
> page the following blurb about Promula under "Known Problems":
> 
> -----------------
> Promula FORTRAN-to-C translator 
> 
> The Promula FORTRAN-to-C translator V4.25 incorrectly handles FORTRAN BLOCK 
> DATA statements.
> Consequently, certain variables in the test program fortran/ftest.f are not 
> initialized to their necessary
> values -- leading to a failue of the make test command in the fortran/ 
> subdirectory. 
> 
> The workaround is to eliminate the BLOCK DATA statement in ftest.f and 
> initialize all the common blocks at
> the top of the program. 
> 
> This problem has been reported to Promula, and there may be a fix for it by 
> now. 
> 
> -----------------
> 
> 
> I don't think this is fair to Promula, which is a two-person consulting
> firm, with very limited distribution of its products.

I can't comment on whether or not the notice in the netCDF `Known
Problems' page is fair or not unless and until we can agree on an
ethical system for judging fairness (otherwise we'll just be talking
past each other).  I do know that the notice is accurate, however.

> There is probably only one version 4.25 in existence for linux.
> Accordingly, this announcement won't be of any use to others, except
> to damage Promula's reputation undeservedly.  Promula reports that
> this problem does not arise with other versions of their compiler.

Yes.  Apparently, later versions of the translator work correctly.
I've been working with George Juras of PROMULA on this issue and he
reports that the newer translators correctly handle the BLOCK DATA
initializations.

> Would you be willing to remove this bug notice? It does not give notice
> of a bug in netCDF, but instead a bug in a compiler.

I'm willing to modify the notice.  The purpose of the `Known Problems'
web page is to reduce the time that we spend responding to client
requests for netCDF support.  Consequently, it's entirely appropriate to
include compiler problems as well as netCDF problems.  You should note
that the page also includes problem notices for the Cray m4(1) utility,
the Cray C++ compiler, and the HP-UX f77(1) compiler.

Please tell me what you think of the modified web page:

    http://www.unidata.ucar.edu/packages/netcdf/known_problems.html

--------
Steve Emmerson   <address@hidden>