Date: Wed, 29 Nov 2017 02:49:02 +0000
From: Taylor R Campbell <campbell+netbsd-source-change...@mumble.net>
Message-ID: <20171129025342.734f860...@jupiter.mumble.net>
| If it turns out that this
| test is legitimately wrong on VAX, we can conditionalize it for VAX.
It wasn't whether it is going to be wrong there, as such, but whether
we can expect a binary value computed on an amd64 (even when generating
vax code) to be bit for bit identical to that which the vax calculates
itself (not using a IEEE floating point format, since that didn't exist
when the VAX float format was originally designed, back in pdp-11 days,
or perhaps even before.)
The same would be true of any other non-IEEE floating implementation.
My first unix computer (the first I worked on, I didn't own it...) was
an Interdata (later Perkin Elmer) system, which used something like
IBM mainframe float formats (the exponent was in powers of 16, not 2).
Also, please do remember that this test is *not* testing floating point,
it is testing locales - the kind of info that you're expecting would,
I think, be better found in a test that is explicitly testing floating
point, and arithmetic, than one which is really making sure that 1234,5678
in a European (mainland) locale parses the same way as 1234.5678 does in
a C/POSIX/UK/US/Aust/Asian locale.
kre
