[v3] annex D 8 and 9 for C++0x

Gabriel Dos Reis gdr@cs.tamu.edu
Sat Oct 27 16:50:00 GMT 2007


On Sat, 27 Oct 2007, Ed Smith-Rowland wrote:

| Gabriel Dos Reis wrote:
| > On Sat, 27 Oct 2007, Ed Smith-Rowland wrote:
| >
| > | math functions.  There might be some API differences in the regex stuff
| > | and
| > | maybe in a few other places.
| >
| > I added literal types supports to the regex library (C++0x).  Last
| > time I hard of it, it might also go through a slight tweak.  But,
| > nothing major.  So the API between TR1 and C++0x will be different.
| > But, nothing major.  And I'm not sure the version in TR1 should be
| > synced with that in C++0x.  After all they are different specs.
| >
| > -- Gaby
| >
| >   
| I agree that it would be a good idea to have tr1 and std be true to the
| respective standards.  It seems like it would be very possible to have one
| representation that would serve both standards with very little pain.  There
| would be some mechanism to support the two as appropriate.  It actually looks
| like someone did this with the tr1_impl directory and the std headers
| including from there.  I didn't see in any of the three regex headers where
| certain features were allowed from tr1 and not std or vice versa.  Maybe when
| std is more ... std there can be switches in the impl to allow differences:
| #ifdef INCLUDE_AS_TR1
| #else
| #endif

I wouldn't like us to suddenly deprecate tr1 stuffs.  I'm all for
mechanisms that would make both specs implemented by single sources if
that is possible.

| 
| About this separate math function standard.  Where would that be?  I'd like to
| check it out.  Also, is there a TR2 brewing somewhere?  All I see is the RFP.

yes, there are tr2 stuff in the pipeline.  for example, the filesystem
library has been carefully sequestered in tr2, to be released after
C++0x :-(

-- Gaby



More information about the Libstdc++ mailing list