Re: RFC: compile with BWX thus only support EV56 and later CPUs
On Sun, May 31, 2015 at 02:56:22AM +0200, Adrian Zaugg wrote:
> It always sad to give up some machines, but if it saves the life of
> others for some time longer, it should be ok. Nobody really objected for
> almost a year to your proposal. I guess all of the persons behind the 12
> alphas counted in the popcon are on this list and at least one would
> have raised the voice if a machines of those was concerned.
I have not done it yet. I have recently managed a local rebuild of
gcc-4.9 and gcc-4.8 that compile with bwx by default but the build
of glibc with those compilers fails with a number of segmentation
errors in the test suite, and until that is resolved I most certainly
will not enact the change at debian-ports.
> You mention it eases your work and accelerates the machines supporting
> BWX CPU flag, so I think it is a good idea to drop support for older
> CPUs and enjoy support for the remaining alpha systems for some time
> longer.
A number of the bugs in the C-compiler were due to support for
legacy ev4, and I suspect some other bugs in the alpha distribution
would disappear if recompiled for BWX. I have too many other projects
I want to work on, that I just don't now have time to put into Alpha
support.
I see that cmake now fails to build, and that means there is a growing
part of the repository that we cannot build. I took a look at the cmake
failure to build but I do not know anything about cmake, and am somewhat
baffled by the build failure. It is a test suite failure in the test
CompileFeatures while testing the "c_function_prototypes" feature
(see [1]), but as far as I can tell that is just a test on the gcc
version and if it is recent enough cmake says you have the feature.
The Alpha gcc does it in fact have the feature. I do not see why
cmake should mistake that. If someone with better C++ and cmake foo
than I have would investigate that would be great!
Cheers
Michael.
[1] http://buildd.debian-ports.org/status/package.php?p=cmake&suite=sid
Reply to: