by jiml8 » Apr 18th, '19, 19:29
It has been my observation that, regardless of intent, support for 32 bit is decaying away.
As much as anything, this seems to be happening because the developers have a dearth of working hardware on which to test, and the toolsets continue to evolve...with the old 32 bit code not being adequately tested as this evolution occurs. When an update to, say, gcc occurs and it breaks 32 bit support, no one notices until way, way downstream. Then, fixing that problem becomes a low priority upstream.
For myself, I am working on a major update to the firmware in the system our company sells, and this involves a migration from FreeBSD 8.4 to FreeBSD 12. Our package has been 32 bit, and I tried to migrate to FreeBSD 12 32 bit in order to support backward compatibility with some of our oldest devices (which are 32 bit) but I found a key feature of FreeBSD 12 to be broken in the 32 bit version. After debating the merits of fixing it (I could, though it is deep in the kernel) vs making the technically right decision of moving to 64 bit, I decided to move to 64 bit, and we will block these very old boxes from trying to update (there's only a few hundred of them out there).
So, I think this problem is beginning to be widespread and it will continue to spread. Rather than a date certain after which there will be no 32 bit support, many projects will just see that support decay away.