Horton

Re: Bug fix for glibc broke freebsd build in REL_11_STABLE

Andrew Gierth <andrew@tao11.riddles.org.uk>

View thread View on postgr.es
From: Andrew Gierth <andrew@tao11.riddles.org.uk>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Andres Freund <andres@anarazel.de>, Victor Wagner <vitus@wagner.pp.ru>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2018-09-04T22:10:12Z
Lists: pgsql-hackers 
>>>>> "Tom" == Tom Lane <tgl@sss.pgh.pa.us> writes:

 >> The reason it behaves oddly is this: on i387 FPU (and NOT on arm32
 >> or on 32-bit i386 with a modern architecture specified to the
 >> compiler), the result of 1e200 * 1e180 is not in fact infinite,
 >> because it fits in an 80-bit long double. So __builtin_isinf reports
 >> that it is finite; but if it gets stored to memory as a double (e.g.
 >> to pass as a parameter to a function), it then becomes infinite.

 Tom> Ah-hah. Can we fix it by explicitly casting the argument of isinf
 Tom> to double?

No; the generated code doesn't change. Presumably the compiler regards
the value as already being of type "double", just one that happens to be
stored in a longer register.

-- 
Andrew (irc:RhodiumToad)

Commits

  1. Error out for clang on x86-32 without SSE2 support, no -fexcess-precision.

  2. Fix 8a934d677 for libc++ and make more include order resistant.