Re: How severe are FTBFS bugs caused by the source using uname?
Theodore Ts'o <tytso@mit.edu> writes:
> On Wed, May 19, 2004 at 12:02:10AM +0200, Goswin von Brederlow wrote:
>> Hi,
>>
>> how severe are FTBFS bugs caused by the source using uname to
>> determine the host/build architecture instead of using
>> dpkg-architecture?
>>
>> The problem is that on multiarch architectures, most notably amd64
>> cpus, the uname will not resemble the host/build architecture:
>> (from a normal Debian Sid i386 system running a 64bit kernel)
>
> Does config.guess do the right thing in this instance? I don't use
> uname -a directly in my packages, but I do use config.guess (like many
> projects that use autoconf). I don't have an x86_64 machine to check
> and make sure autoconf does something sane on those architectures; I
> presume that it does, hopefully?
We haven't tested compiling i386 debs on amd64 on a large scale yet
but only tested those that failed under 64bit already. Of those
'broken' packages a percentage are broken for 32bit too for all thats
worth.
What does the 'normal' config.guess use? If your package detects
i686-linux or similar instead of i486-linux you might have the same
problem.
The multiarch port is currently starting up again and their 32bit
autobuilder should find all those errors over time. So just sit back
and wait for a bugreport. :)
MfG
Goswin
Reply to: