This is the mail archive of the gcc-patches@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]
Other format: [Raw text]

Re: [gfortran] patch for PR libfortran/25425


:REVIEWMAIL:

On Mon, Feb 06, 2006 at 10:28:02PM +0100, FX Coudert wrote:

> Attached patch fixes PR libfortran/25425: list-directed output of 0.0 
> changed between F95 and F2003. The former required an E format to be 
> used, while the latter requires a F format.

That's good.

> 
> Bootstrapped and regtested on i686-linux. OK for trunk?

> Another question is: the patch for that issue requires a small change 
> in the set_std() libgfortran function. In the future, this might be 
> usefull in the 4.1 branch (for this or other patch which introduce an 
> explicit dependency in the library on the standard chosen, not only for 
> warnings and error messages). Can this be done after the release of 
> 4.1.0? Or is it forbidden by binary compatibility, in which case we 
> might want to push it a bit so that it's present in 4.1.0?

First, I agree that standards warning in the library should only
be given when "-pedantic" is set.  Could you also mention this
in the documentation?

Second, this change is something we can't do between 4.1.0 and 4.1.1.
Adding another argument to a function will break things when there
is a mismatch between compiler and library.

It's probably too late now, but for future cases, a bit mask would
probably be better; it can be extended without many problems.
When we redesign the whole library interface for 5.0, we can try
once more :-)

So, I'd say this is OK trunk, and for 4.1 a bit later (to get
this in before the release), unless somebody speaks up against
this within 24 hours or so.


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