Anand Avati <
address@hidden> wrote:
> Not "instead", but along. syncproc is a pthread which executes synctasks
> (and syncops). So a synctask_set() performed in one syncproc will not be
> obtained via synctask_get() performed in another (original) syncproc. So
> instead of NULL we could get an unexpected (and maybe free'd/corrupted?)
> synctask pointer. If we either avoid bouncing of synctasks between
> syncprocs, or limit syncenv to a single syncproc, then your patch will be
> "complete".