Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 12 Apr 2005 00:18:07 -0600
From:      Scott Long <scottl@samsco.org>
To:        Michael Nottebrock <michaelnottebrock@gmx.net>
Cc:        Mikhail Teterin <mi@corbulon.video-collage.com>
Subject:   Re: mozilla's install hanging on amd64
Message-ID:  <425B681F.5090506@samsco.org>
In-Reply-To: <200504120805.23933.michaelnottebrock@gmx.net>
References:  <200504120533.j3C5XNFL008134@corbulon.video-collage.com> <200504120805.23933.michaelnottebrock@gmx.net>

next in thread | previous in thread | raw e-mail | index | archive | help
Michael Nottebrock wrote:
> On Tuesday, 12. April 2005 07:33, Mikhail Teterin wrote:
> 
> 
>>>Mikhail seems to have forgotten that not too long ago, CPUTYPE was
>>>something for only the most daring would use. GCC's optimizers have seen
>>>much improvement since then, but just because you can get away with
>>>always setting CPUTYPE for everything much more often these days doesn't
>>>mean it's not risky anymore (or we would have the resources to runtime
>>>test every port in the collection with all possible CPUTYPE settings on
>>>each arch).
>>
>>Nothing except Mozilla has ever caused problems for Mikhail, that was
>>traceable to this switch.
> 
> 
> Lucky Mikhail. I actually don't believe you really missed the tons of broken 
> ports that -march=p4 used to produce with early versions of gcc3 though.
> 
> 
>>make.conf(5) documents it, it should work. Period.
> 
> 
> make.conf(5) documents CFLAGS. What would you like to infer from that fact?
> 
> 
>>And everything does 
>>work. Complex things like Perl build fine and pass their self-tests (make
>>test). The entire KDE built and works (although it is lacking self-tests).
>>
>>Time to stop blaming compiler for the software's bug -- and Mozilla has
>>plenty of them.
> 
> 
> If a compiler optimization produces a bad binary while the same compiler with 
> the switch off does not (or a different version of the compiler with the 
> switch does not), the compiler usually *is* to blame. I don't know of course 
> whether this is the case, I couldn't find the beginning of this thread.
> 

This isn't always true.  Pointer aliasing rules become stricter but
still spec-legal with certain optimizer modules, for example.  I'd 99%
bet that mozilla or a dependency is to blame here.  However, that
doesn't mean that the port is non-functional.  It means that it has
known bugs in the non-default case that should be documented until they
are fixed.  I'm glad that Mikhail has exposed one of these problems, and
I hope that both a temporary and permanent solution can be found.

Scott



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?425B681F.5090506>