question about sign of DFP interest

[Joseph Myers]

On Thu, 12 Feb 2026, Steve Summit wrote:

> So is there any sense that DFP support of some flavor definitely
> does not belong in glibc?  Or are people just not interested in
> the question?  (I suspect that if there are truly no interested
> parties, it may be difficult to get anyone to review any
> candidate DFP code, but I'll hold that concern for another day.)

Who would use the code?  What widely used free software packages in 
GNU/Linux distributions (for example) use DFP (for functionality that is 
itself widely used, not just e.g. providing optional C library bindings in 
another language)?  Who would maintain the code long term?  Who would 
write and maintain all the surrounding infrastucture long term, such as a 
correctly-rounding DFP analogue of all the MPFR interfaces used in 
gen-auto-libm-tests so that we can avoid duplication of test 
infrastructure and share test inputs between binary and decimal FP tests 
(the numerical values of the test inputs and the logic for turning those 
into actual floating-point values embodies a lot of information about 
which cases are harder to get right in implementing <math.h> functions, 
which it's important to reuse in DFP testing - although those inputs that 
reflect edge cases related to precision of different input types, e.g. 
where the largest exactly representable odd integer is involved, would 
need corresponding tests added for the DFP formats as well)?  Even a 
minimum viable product for DFP support would involve large amounts of 
infrastructure to support such testing (shared with the binary FP test 
support) at all.

-- 
Joseph S. Myers
josmyers@redhat.com