From: Doug Gregor (gregod_at_[hidden])
Date: 2001-04-17 15:23:47
On Tuesday 17 April 2001 04:15, you wrote:
> The new version of CRC I just now downloaded from the vault may have set a
> new record. At least I can't remember a single test program that caused
> internal compiler errors from TWO of the Win32 compilers.
> First try, Borland got an internal compiler error, Metrowerks compiled OK,
> but the others failed. (See First try, below.)
> I commented out "typename" at the indicated places, and tried again. Now
> GCC joined Borland in the internal compiler error department. Metrowerks
> continued to compile without error, while Microsoft reported errors (see
> Second try, below)
> It is really hard to deal with heavily templatized code that hasn't been
> checked out by at least two compilers:-(
Perhaps the formal review of a library should be prefaced by a few days where
other Boost members could try the library on several different compilers
before the library actually comes up for formal review. At this point the
library would be slushy - internals may need to be modified to make the code
portable, but the interface should be unchanged.
I believe this would make formal reviews go more smoothly, and allieviate the
problem with libraries changing while they are being reviewed.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk