Reply to INFO-ADA Digest V93 #184

[mvensky@nosve.elan.af.mil]

Date: 29 Mar 93 10:44:34 EST
From: Larry iummo <75570.2443@CompuServe.COM>
Subject: Rep clause problems on VADS
Message-ID: <930329154433_75570.2443_EHA28-1@CompuServe.COM>

We're using the Verdix (Sun) Ada compiler on a SPARCstation and VADSwork
68030 cross-compiler targeted to a MVME-147/143 network sending
inter-processor messages via sockets between the SPARC and the 68k netwok.
We send quite a few variant records across the network that we have used
rep clauses in to align the fields properly.  It work fine on message that
do not containt variantparts.When variant parts are used, the 68k seems
to ignore the designated byte layout specified by the rep clauses.  We haheard 
t
hat they(the "at" clause) may be ignored on the VADSworks side.
-----------

I have been laboring mightily trying to read in FORTRAN binary files using
variant records and  SEQUENTIAL_IO. I have found considerable variation between
compilers as far a supporting this use of SEQUENTIAL_IO. CDC's Ada on NOS/VE
can't even read records back after it has written them. Whereas Meridian on
the Mac, Sun Ada on a SparcServer 690MP, and CDCs version of Ada on it 4000
series all can read/write variant records using SEQUENTIAL_IO.

The Ada lawyers at CDC say the LRM does not require them to support variant
records with SEQUENTIAL_IO. My reading of the LRM concurs, however I think
it is a best ambiguous and at worst chicken-shit for CDC to take this stance
in light on how the other Ada compilers respond.

I think you are having a similar problem with this particular construct. I woul
d
be interested to know how it turns out.

Mvensky@nosve.elan.af.mil