http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34928
--- Comment #9 from Bud Davis <bdavis at gcc dot gnu.org> --- I completely support closing this PR with a note in the documentation. On shared memory mini computers of a bygone era, it was common to map the common blocks to a specific memory address, and then more than one program (in more than one cpu) could access them. You used the volatile keyword to ensure the writes happened to memory, and didn't stay in the register set of a cpu. I also wouldn't be surprised if I/O devices (think VME) were mapped to a fortan common block name, and thus also required volatile. Both concepts are long gone, but the code still lives. A note in the documentation is all it needs. --bud p.s. This explanation may be wrong. It was a long time ago. !!
