Re: [RFC] Large file iostreams

[Benjamin Kosnik <bkoz at redhat dot com>]

>..and there doesn't seem to be a way to test that streamoff is not a
>typedef, and instead a type via class defined in std. Why should we
>bother with the __streamoff bits when it can be class streamoff? 

.... and as Nathan has indicated, we can just typedef the class anyway.

>>The attached patch passes the testsuite. It differs from the last
>>version in that operators ++ and -- have been added to streamoff,
>>to make testsuite/27_io/objects/char/10.cc pass. 

Let's remove that bit for the time being to see if we can get away without them.

>Anyway. Let's move streamsize and streamoff into postypes.h, as well as
>streamoff_base_type. See the patch I sent last night. That way all these
>types are consolidated into one file.

Done.

I've attached the patch that I checked in.

best,
benjamin

Attachment: 20031016-3.patch.bz2
Description: BZip2 compressed data