From: Iain K. Hanson (iain.hanson_at_[hidden])
Date: 2005-05-04 09:56:08
On Wed, 2005-05-04 at 13:19 +0300, Peter Dimov wrote:
> Beman Dawes wrote:
> > "Peter Dimov" <pdimov_at_[hidden]> wrote in message
> > news:000e01c54fdb$be752eb0$6401a8c0_at_pdimov2...
> >> I also don't like the fact that the error codes aren't the standard
> >> errno E* constants, but we already had this debate once. :-) (As
> >> with threading, it is my opinion that Posix should be acknowledged.)
> > The intent _is_ to supply the actual system error code (errno for
> > POSIX). There is another function available to convert to a portable
> > code if desired.
> My point was that errno should be the portable error code. This will save
> you one mapping. Your arguments for not using E* were purely
> implementation-driven and at this stage of the library development it makes
> sense to revisit the issue from standardization point of view.
I agree that it is time to re-visit these issues and they are going to
apply to more than just the Filesystem library. I don't believe The E*
names are portable to non-posix systems which includes ( at least part
of ) Windows.
I think that by default, C++ libraries should return C++ error names and
not POSIX or any other platform or native error names. There should
probably be a convention for mapping between exception names and error
names. This will ease the programmers task as the move between
just my 2p worth.
This email has been scanned for all known viruses by the MessageLabs Email
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk