develooper Front page | perl.perl5.porters | Postings from November 2000

Re: sfio97 vs sfio2000

Thread Previous | Thread Next
Nick Ing-Simmons
November 15, 2000 09:14
Re: sfio97 vs sfio2000
Message ID:
Nicholas Clark <> writes:
>How many versions of sfio are we supporting?

Pass. It is rather un-interesting to me personally.
It is not a major speed win on Solaris/Linux, and if it was
I would rather fix "perlio" to do the same things.
And I don't like the license muddle.

>> The recent change away from sftell() does _NOT_ work right with sfio97.
>> As it was only to give error rather than something that _might_ be useful
>> I have reverted that to use sftell().
>I was in two minds about whether it should change.
>status quo:	it tells you something useful
>change:		I was under the impression that perl scripts shouldn't notice
>		anything if sfio is substituted for stdio
>now that Jarkko says that C's ftell()'s result is differs for ttys on
>different platforms I'm coming round to the view that perl's tell using
>sftell is best.

That is what it does at the moment.

>> Also the perlsfio.h that I inherited seems to have turned off 
>> PerlIO_canset_cnt() - and sfio _can_ set count same way that Linux/glibc
>> stdio can - set the pointer and count comes too.
>Assuming it was me I was attempting to distinguish can set count
>(count is a lval), and count changes as a side effect of setting ptr
>I felt that this was important. 

It is not important as far as I can see - and as I designed both 
the original PerlIO stuff and now the new stuff I assume I can 
spec' it to be not important...
Any code that sets count needs to set ptr as well - seeing as 
Linux, sfio and now perlio all do it that way.

>I don't think it was documented before,
>and I am guilty of not attempting to rectify that.
>Nicholas Clark
Nick Ing-Simmons <>
Via, but not speaking for: Texas Instruments Ltd.

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About