On Fri, Sep 25, 2015 at 11:54 AM, Arun Raghavan <arun at accosted.net> wrote: > On 25 September 2015 at 15:21, Tanu Kaskinen <tanuk at iki.fi> wrote: >> On Thu, 2015-09-24 at 23:21 +0200, Ahmed S. Darwish wrote: >>> OK, after a second look, it seems that the anonymous structure indeed >>> does not provide any extra benefit. So I believe everyone now agrees >>> now that the following definitions are the most appropriate: >>> >>> typedef struct { >>> pa_mem mem; /* Parent; must be first */ >>> int fd; >>> } pa_memfd; >>> >>> typedef struct { >>> pa_mem mem; /* Parent; must be first */ >>> int id; >>> bool do_unlink; >>> } pa_shm; >>> >>> Good :-) >>> >>> But I can't get my head around not using the anonymous unions, and >>> basically whether they provide any perceived disadvantage: >> >> I don't think Arun objected to the use of anonymous unions. I certainly >> don't see anything wrong with using anonymous unions in pa_mempool. > > Yup, and I can see why I might have come off that way. I think this > pattern is fine. > Great, thanks a lot everyone. -- Darwish http://darwish.chasingpointers.com