Guillem Jover wrote:
Resulted printouts should be:
$ dpkg-architecture
[...]
DEB_BUILD_GNU_CPU=i486
Why is this different than the one on GNU_TYPE?
DEB_BUILD_GNU_SYSTEM=solaris
This should match the GNU_TYPE as well, so solaris2.11.
DEB_BUILD_GNU_TYPE=i386-pc-solaris2.11
The -pc- part should not be here.
DEB_HOST_GNU_CPU=i486
DEB_HOST_GNU_SYSTEM=solaris
DEB_HOST_GNU_TYPE=i386-pc-solaris2.11
The same applies for this ones.
One of the issues we originally encountered is the artifact
associated with SunOS 5.x (Solaris) ports of many OSS packages,
in that they follow the following scheme:
{i386,sparc}-{sun,pc}-solaris2.xx
Where:
i386-pc-solaris2.xx is for x86/x64 platforms
sparc-sun-solaris2.xx is for SPARC platforms
xx is `uname -r | sed -e 's/[^.]*//'`
For example, gcc/gcj/binutils config.guess implements such
a logic. In many instances, however, dpkg-architecture output
is used instead to override config.guess. When this is done
with the default i486-solaris (for $DEB_HOST_GNU_TYPE), all
of the Solaris-specific code end up getting disabled.
Our proposal allows for these OSS packages/programs to compile
seamlessly given the fact that dpkg-architecture provides the
"correct" OS/platform string to match config.guess's logic.
One thing that I would like to add/correct with respect to
our original proposal, is that the solaris2.11 string is not
hard-coded and instead derived from uname (see "xx" above)
at dpkg-architecture runtime. This should be addressed in
the final patch.
If you think this is horribly broken, please advise us on
an alternative (and the least invasive) solution.
Thanks,
--Mac
--
Nexenta Systems, Inc.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]