This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: g77 0.5.23 and higher, egcs 1.1.1: Changed behavior


Roger Hill-Cottingham wrote:
> 
> Toon Moene wrote:
> >
> > Michael Lemke wrote:
> >
> > > I have a question about the newer versions of g77, especially as it
> > > comes with egcs 1.1.1.  Why has the error level of
> > >
> > >         OPEN( ...., READONLY, SHARED )
> > >
> > > been changed from warning to fatal?  This is causing me a lot of code
> > > changes and makes the code incompatible with the VMS version.  Before
> > > the change I could just leave the code in and it would compile on VMS,
> > > g77 and Solaris.  Now that.  What's the reason?
> >
> > Hmmm, I wasn't aware that this had changed ...
> >
> > > P.S.  I would appreciate an e-mail cc.
> >
> > We too - to egcs@egcs.cygnus.com :-)

Done. 

> >
> > [ Forwarded ]
> > > --
> > > Michael Lemke
> > > INA BusinessFlow Administration (KOU2, Tel. 3441)
> > > lemkemch@ina.de
> >
> > --
> > Toon Moene (toon@moene.indiv.nluug.nl)
> > Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
> > Phone: +31 346 214290; Fax: +31 346 214286
> > g77 Support: fortran@gnu.org; egcs: egcs-bugs@cygnus.com
> 
> Yep, it changed after egcs-1.0.3.  In ...gcc/f/ChangeLog you will see:
> 
> Tue May 19 14:52:41 1998  Craig Burley  <burley@gnu.org>
> 
>         * bad.def (FFEBAD_OPEN_UNSUPPORTED, FFEBAD_INQUIRE_UNSUPPORTED,
>         FFEBAD_READ_UNSUPPORTED, FFEBAD_WRITE_UNSUPPORTED,
>         FFEBAD_QUAD_UNSUPPORTED, FFEBAD_BLOCKDATA_STMT,
>         FFEBAD_TRUNCATING_CHARACTER, FFEBAD_TRUNCATING_HOLLERITH,
>         FFEBAD_TRUNCATING_NUMERIC, FFEBAD_TRUNCATING_TYPELESS,
>         FFEBAD_TYPELESS_OVERFLOW): Change these from warnings
>         to errors.
> 
> I have applied the following to my copies of subsequent versions to
> revert to the original behaviour:
> 
> tesla:/usr/local/src/egcs-1.1b/gcc/f $ diff -u bad.def.orig bad.def
> --- bad.def.orig        Tue Sep  8 10:59:06 1998
> +++ bad.def     Tue Sep  8 10:59:29 1998
> @@ -551,7 +551,7 @@
>  "Unsupported FORMAT specifier at %0")
>  FFEBAD_MSGS1 (FFEBAD_FORMAT_VARIABLE, FATAL,
>  "Variable-expression FORMAT specifier at %0 -- unsupported")
> -FFEBAD_MSGS2 (FFEBAD_OPEN_UNSUPPORTED, FATAL,
> +FFEBAD_MSGS2 (FFEBAD_OPEN_UNSUPPORTED, WARN,
>  "Unsupported OPEN control item at %0 -- ACTION=, ASSOCIATEVARIABLE=,
> BLOCKSIZE=, BUFFERCOUNT=, CARRIAGECONTROL=, DEFAULTFILE=, DELIM=,
> DISPOSE=, EXTENDSIZE=, INITIALSIZE=, KEY=, MAXREC=, NOSPANBLOCKS,
> ORGANIZATION=, PAD=, POSITION=, READONLY=, RECORDTYPE=, SHARED=, and
> USEROPEN= are not supported",
>  "Unsupported OPEN control item at %0")
>  FFEBAD_MSGS2 (FFEBAD_INQUIRE_UNSUPPORTED, FATAL,
>

Yep, I had found that place, too.  Also the note in the change log
file.  But before I do the same patch and create yet another
incompatibility (my g77 vs the rest of the world) I want to know the
rationale behind this change.  The explanation in the change log file is
not sufficient.  I can understand a fatal error for stuff like USEROPEN,
KEY, OPRGANIZATION='KEYED' but not for READONLY, SHARED and probably a
few others.  

If this is not going to change provide an option to turn this into a
warning message.

Why do new versions always break compatibility unless it's VMS?

Michael
-- 
Michael Lemke
INA BusinessFlow Administration (KOU2, Tel. 3441)
lemkemch@kou1.ina.de


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]