Re: Gcc bugs break java/jdk15 build? [Workaround]

From: M. Warner Losh <imp_at_bsdimp.com>
Date: Wed, 22 Aug 2007 20:18:51 -0600 (MDT)
In message: <20070819073953.GA36594_at_Gatekeeper.FreeBSD.org>
            Stefan Eßer <se_at_freebsd.org> writes:
: On 2007-08-19 12:01 +0900, Ken Yamada <ken_at_tydfam.jp> wrote:
: >   Does this mean that GCC 4.2.1 does not solve "loop optimization bug" 
: >   pointed out by Andrey Chernov (http://docs.freebsd.org/cgi/getmsg.cgi?fetch=282888+0+archive/2007/freebsd-current/20070701.freebsd-current) on 4.2.0?
: >   
: >   Is it a good idea of adding -fno-tree-vrp to CFLAGS in /etc/make.conf 
: >   to avoid this optimization bug for all compilation?  (BTW, I leave -O3 
: >   of CFLAGS as is in make.conf and just added -fno-tree-vrp to the end of 
: >   the line.)
: 
: I did just that (added -fno-tree-vrp to CFLAGS) after it was found
: that gcc-4.2 had the optimizer bug, and have continued to build my
: system with -O2 (or other optimization levels as required). This 
: seems to work just fine (no problems with ports that are not working
: when built with a plain -O2). I had suggested to make -fno-tree-vrp
: the default for our CC (could be overridden on the command line, but
: would allow to use all -O levels without being hit by the compiler
: bug; but it appears that nobody else liked the idea at that time, 
: i.e. no response at all ...).

Actually, it was added to the tree and then backed out because it
broke the upgrade path.

Warner
Received on Thu Aug 23 2007 - 00:19:50 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:16 UTC