Re: mingw isnanl-nolibm failure

[Eric Blake]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

According to Bruno Haible on 12/7/2007 4:19 AM:
> 
> Either this, or ensure that the "checking where to find the exponent" tests
> return the bit positions instead of "unknown".
> 
> But I don't see how to implement either of these two possible fixes for a
> cross-compiler, in a way that works also for other platforms than mingw.
> Any ideas?

How expensive is a run-time test to find exponent positions, and how
likely would such a test be to run without causing a reference to an
invalid pattern?  Maybe the solution is to code a run-once check into the
cross-compiled app with unknown configure-time exponent position, so that
the rest of the code can fall back on a computed instead of compile-time
value during execution.  Yes, that would mean that cross-compiled
performance is a bit slower, but at least it would be generically correct.

As for configure-time cross-compiling checks, could we write up a test
similar to the word-endianness checks, where known sentinels are injected
into the .o file via structure initialization to allow examination of the
layout of that .o file?  That is, come up with something like this
untested snippet:

struct
{
  char pre[16];
  double value;
  char post[16];
} checker { "pattern1", 1.0, "pattern2" };

then grepping through the object file for the two patterns and observing
the bit pattern of the floating point value in between?

- --
Don't work too hard, make some time for fun as well!

Eric Blake             address@hidden
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Cygwin)
Comment: Public key at home.comcast.net/~ericblake/eblake.gpg
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHWUZF84KuGfSFAYARAr+aAJ0b9CiIb/2BxNBeOGmzDD39VC/31wCdFUO8
/7n7fRPqYqWSjriAVYH/y1o=
=vOC7
-----END PGP SIGNATURE-----