Re: `` redirection bug

On Sat, Oct 28, 2000 at 08:10:18PM +0100, Richard Proctor wrote:
> On Sat 28 Oct, James Taylor wrote:
> > PS. Is anyone working on a port of Perl 5.6 ?
> > 
> 
> YES.
> 
> I am working on getting the existing build to rebuild on my machine and as
> soon as I have I intended to go to the leading edge of perl.  It is possible
> I will skip 5.6...  I am while typing this actually downloading the current
> bleeding edge perl to compare with the current build.
> 
> As soon as I have it "working" I wont to spend a little time closing the gap
> between riscos perl and the main perl development, to make it much easier to
> keep riscos perl in line.
> 
> I am having several small problems, but progess is real.

I went to YAPC::Europe and while there had a chat with Jarkko Hietaniemi
about merging the RISC OS port back into the main tree. Things sort of
stalled about 2 years ago on this. One thing he doesn't like is the way
thing like os_file are called with magic numbers (such as 5) - and was
expecting symbolic constants (of the sort defined in header files)
I showed him a copy of the RISC OS PRM to demonstrate that they *are* names
not numbers :-)
Currently I think the best compromise (one certainly acceptable to me and
to him) was to #define all the numbers needed for things like OS_File,
OS_GBPB etc in one place (presumably riscosish.h). I was intending to
cut&paste the definitions used by OSLib, but was intending *not* to make
perl dependent on anther library's headers.

I'm about to see if I can get sfio2000 ported (sfio98 was mostly done about
2 years ago) as it's better than UnixLib stdio, and allows more reliable
sticking of libraries in zip files. Which is probably going to be wanted
as an option for size reasons (as well as long filenames and 77 directory
entry limits) as with Unicode current bleading edge perl has 7408522 bytes
of files in the lib directory.

Personally I'm not using 5.6 for production purposes (on Unix) because
rather too much of Unicode support is a mess, (and it's not an compile
time option to disable it). 5.6.1 should be considerably better (but
not perfect), but 5.6.1 doesn't seem to be very happening.
[5.6.1 is effectively going to be bug fixes back-patched from the 5.7
tree]

Nicholas Clark