Endian[n]ess, language-specific problems and help with egcs.

[Toon Moene <toon at moene dot indiv dot nluug dot nl>]

Lectoribus Salutem,

Recently, we've had some uproar related to the questions of one  
Fortran user of egcs with respect to the treatment of unformatted  
sequential files between his normal working environment (an SGI or  
HP workstation) and Linux on a ix86 PC with egcs.

As we have experienced, it is hard to:

1. Pin down that this is an endian[n]ess problem.
   (Initially, the question was about 32-bit IEEE floating point
    formats being different between Linux PC's and "workstations")

2. Pin down that the problem really boils down to the question of
   how to convert Fortran sequential unformatted files from one
   endian[n]ess to another.

3. Stop the stream of well-meant but not-completely-correct to
   utterly-useless advise by people who misunderstood the original
   problem.

I am not going to furnish a solution for this - I don't have one.

I do think some simple advise might be helpful in these cases:

1. As soon as the term "fortran" or "g77" occurs, PLEASE REFRAIN
   FROM REPLYING.  Craig, Dave, Kate, Mumit or myself will guide
   the question(er) in providing more information to fortran@gnu.org
   the mailing list meant to deal with GNU Fortran questions.

2. Although the question posed might have an answer outside of the
   realm of programming languages - don't force it upon the original
   questioner - s/he might not understand the full implications of
   the ordeal s/he has to go through, and it won't help showing it
   all to them.

3. Refrain from making snide remarks, even if not directed at the
   questioner.  A problem like endian[n]ess is a showstopper for
   a Fortran programmer, while it's a part of life for the
   experienced C programmer (htons and all that).

However, reviewing the posts of the last two days, it occurred to  
me that Zhu Min had little choice than to put his question to  
egcs@cygnus.com - there is a egcs-bugs@cygnus.com, but it was not  
clearly (and in fact wasn't) a _bug_ in egcs; note that there is no  
egcs-help@cygnus.com.

Improving FAQs only goes so far.  The problem every programmer  
stumbling over a Segmentation Violation (and I'm quite sure that  
this was what Zhu Min experienced trying to read an unformatted  
sequential file in the wrong endian[n]ess) faces is:

A. Is this a problem in my code ?  But it worked perfectly on (set
   of machines);
B. If it is not, what changed in my environment ?  Uh oh, I used
   another compiler.
C. Egcs, isn't that *E*xperimental GNU Compiling System ?

Thanks for your attention.
Toon.