[RFC] preliminary patch to move decimal float runtime to new libs

[Joseph S. Myers]

On Thu, 14 Aug 2008, Janis Johnson wrote:

> I've made progress on moving decimal float runtime support from libgcc.a
> to libgcc_dfp.a and building a new shared library, libgcc_dfp.so.1.
> The patch works sometimes but still has problems and so is not yet ready
> for a full review, but I'd appreciate comments on the general approach.
> I'll be out of touch until Tuesday.

Two questions:

* Do you use symbol versioning in the new library, and with what symbol 
versions if so?  I think it should probably use symbol versioning.

* Do you *really* need to put configuration variables for this in 
gcc/config/ instead of libgcc/config/?  We have an incomplete transition 
of libgcc configuration from gcc/config/ to libgcc/config/.  In most cases 
it should be possible to transition a configuration variable or group of 
related configuration variables with changes to all configs that define 
those variables - for some variables this may be a lot of work, for some 
not so much.  When adding new variables, I'd really encourage putting them 
in files in libgcc/config/, referenced from libgcc/config.host, rather 
than adding more variables to the legacy list of variables whose values 
are passed from the gcc directory to the libgcc directory.

-- 
Joseph S. Myers
joseph@codesourcery.com