Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 19 Jan 2009 09:12:36 -0800
From:      Maxim Sobolev <sobomax@FreeBSD.org>
To:        Sam Leffler <sam@FreeBSD.org>
Cc:        svn-src-head@FreeBSD.org, svn-src-all@FreeBSD.org, src-committers@FreeBSD.org
Subject:   Re: svn commit: r187426 - head/sys/amd64/conf
Message-ID:  <4974B484.7030608@FreeBSD.org>
In-Reply-To: <4974ABCC.7000107@freebsd.org>
References:  <200901190710.n0J7ACSg001385@svn.freebsd.org> <497432A1.9060805@samsco.org> <497446D4.5020104@FreeBSD.org> <4974ABCC.7000107@freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Sam Leffler wrote:
> Maxim Sobolev wrote:
>> Scott Long wrote:
>>> prepare to be wrong.  And above all else, don't put drivers into here
>>> that you haven't tested.  It's pretty silly to admit in your commit
>>> message, for all to see, that you are blatantly committing without
>>> testing.
>>
>> Actually this is interesting point, what the best strategy for us as 
>> the project should be? Should we new put drivers that have been tested 
>> on i386 only and don't have any particular reason to be i386-specific 
>> (i.e. ISA/EISA drivers, PCMCIA drivers etc) into amd64 GENERIC 
>> automatically and wait for somebody to report a problem, or stay on 
>> the safe side and enable drivers on amd64 only after somebody actually 
>> has tested them and confirms that they are working? Should this policy 
>> depend on driver class (for example a storage driver has much higher 
>> potential for screwing user's data compared to a network driver or a 
>> sound driver) and on release (HEAD / STABLE)? IMHO FreeBSD could 
>> benefit by putting at least non-storage untested non i386-specific 
>> drivers into amd64 kernel and/or at least in HEAD to give them testing 
>> and a wider exposure.
>>
>> This is not just idle interest for me - recently our company has 
>> started shipping amd64 version of our FreeBSD-based product, so that 
>> we are a little bit concerned about hardware support with amd64 7.1 
>> kernel being a little bit narrower compared to i386 7.1 kernel.
>>
>> I apologize if this topic has been discussed somewhere already.
> 
> I think the answer to your question about default-enabling drivers is 
> very clear: it is the decision of the person maintaining the driver.  If 
> you're willing to SUPPORT a driver on a platform then feel free to 
> enable it.  Otherwise doing a drive-by to enable a driver that may or 
> may not work may easily result in complaints that are unanswered.  These 
> have resulted in people concluding wider breakage that easily becomes 
> de-facto and are hard to kill given that people google for help, find 
> old complaints, and stop searching.

OK, makes sense.

By the way, there is a question on this topic to you. The wi(4) has been 
removed from i386 GENERIC, but it is still present in amd64 GENERIC. Is 
it intentional or just a mistake?

-Maxim



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4974B484.7030608>