Re: ISO C++ and Octave

[Mumit Khan]

On Tue, 30 Jan 2001, John W. Eaton wrote:

> On 30-Jan-2001, Mumit Khan <address@hidden> wrote:
> | 
> | 1. Just a few remaining missing namespace qualifiers, a minutes fixed
> | those.
> 
> OK, if you have a patch, please send it on.

Will do, but it'll be a few days before I get to it again.

> | 2. The code for equal (src/mx-inlines.cc) currently causes a problem 
> | for gcc-2.97/libstdc++-v3, but I haven't been able to reduce the testcase 
> | yet. Easy workarounds: (1) use std::equal, which works with older 
> | compilers as well, (2) explicitly use ::equal. It is not a problem in
> | Octave code as far as I can see.
> 
> I'd also be willing to rename those functions to be mx_XXX to avoid
> this kind of problem.

Easily done. I'll add that as a separate patch from (1).

> | 3. This is the sticky one - the interface to the streams has changed
> | significantly and so we do need to update anything built on top of
> | std::filebuf. 
> 
> If I can build a current snapshot of gcc 3.0 maybe I can get an idea
> of how hard this would be to fix.

As of about a week or ago, libstdc++-v3 actually somewhat works on Linux,
at least any program linked against the library doesn't crash at startup.
It's been stable on Solaris 8 for a while now.

> Do you think it would be possible to write code that would work with
> both old and new compilers in these cases?  If not, then the idea of
> using a macro seems OK, or we can hide the details in some function(s).

For my own code, I had no choice but to use the macro to separate old
from the new, but I'll have to understand the Octave implementation more
before I know what will work. From a quick look, it's a question of just
a few changes, but that's being over-optimistic.

I'll dig into it this weekend, and hopefully have an implementation, or
at least a much better idea.

Regards,
Mumit