On 07/15/17 22:35, coypu%sdf.org@localhost wrote:
On Sat, Jul 15, 2017 at 11:22:54PM +0200, Joerg Sonnenberger wrote:It is not reasonable for -D_FORTIFY_SOURCE to add warnings for things within the normal interface contract. If it throws a warning for incorrect use, that's a different thing.It won't do it any more if he updates glibc, but asking someone to update his entire libc to use pkgsrc is silly.
Upgrading the primary libc would break compatibility with some commercial software, which is one of the major reasons people in HPC (High Performance Computing, or supercomputing) choose enterprise Linux over Ubuntu, etc.
We could install a newer libc alongside it in a non-standard location, but this seems like a nasty can of worms.
Installing the latest open source on RHEL/CentOS consumes a lot of man-hours without a tool like pkgsrc, which provides the newer compilers and basic libraries and build tools needed. One alternative for HPC clusters is to partition the resources, running enterprise Linux for commercial apps on some compute nodes and a bleeding edge OS on others for open source. This is costly and hinders utilization, so almost nobody does it. Many are using VMs or containers, but this is also adds a lot of complexity and overhead.
Pkgsrc makes it really easy to install the latest open source alongside commercial apps.
If we're willing to tolerate a few less-than-ideal patches to support enterprise Linux, we can have a big impact on scientific research. We'll also be able to recruit and retain packagers who have no choice but to use enterprise Linux systems, which will help everyone else in the pkgsrc community.
As I've stated recently on this list, the HPC community is starting to warm up to pkgsrc, so now would be a bad time to start adding hurdles for RHEL/CentOS users.
Thanks,
JB
--
Earth is a beta site.