From: Rene Rivera (grafikrobot_at_[hidden])
Date: 2007-10-28 16:09:10
David Abrahams wrote:
> on Sun Oct 28 2007, Vladimir Prus <ghost-AT-cs.msu.su> wrote:
>> On Saturday 27 October 2007 06:21:22 Boris Gubenko wrote:
>>> Index: build.jam
>>> --- build.jam (revision 40483)
>>> +++ build.jam (working copy)
>>> @@ -175,7 +175,7 @@
>>> : -L$(--python-lib) -l$(--python-lib) ;
>>> ## GCC 2.x, 3.x, 4.x
>>> toolset gcc gcc : "-o " : -D
>>> - : -pedantic
>>> + : -pedantic -fno-strict-aliasing
>>> [ opt --release : [ opt --symbols : -g : -s ] -O3 ]
>>> [ opt --debug : -g -O0 -fno-inline ]
>>> -I$(--python-include) -I$(--extra-include) -Wno-long-long
>> given that we have no idea if this gcc bug, or some subtle standard
>> violation on jam part, and how to workaround it property, can we
>> just apply this patch?
>> This gcc 4.2 issue is recurring problem on IRC.
> It's *highly* likely that it's a problem in jam. 'C' code routinely
> uses pointer type-punning with the assumption that the compiler
> doesn't care, but the compiler implements optimizations based on the
> rule that (with a few exceptions) two pointers of different types
> cannot point into the same memory block.
Yep, and I *really* want to know where in bjam the problem is. Because
fixing it might also fix the problems it has with using GC. But I will
add the gcc option since I have failed to find where the bug is, and no
one else has pointed out where the problem is.
-- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim - grafikrobot/yahoo
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk