On Thu, 19 Feb 2009 11:18:56 -0800 Jonathan Adams wrote: > On Thu, Feb 19, 2009 at 07:37:32PM +0100, Roland Mainz wrote: > > AFAIK this will not work properly if someone uses libumem via LD_PRELOAD > > since libumem&co. are not async-signal safe and ksh93 relies on this > > feature.
> I asked this before with no answer, but how does ksh93 do this without a > substantial performance overhead? its passive a signal handler function can detect if any ast malloc routine is currently interrupted by the handler the busy condition test is (vmstat(0,0)!=0) provided the <vmalloc.h> api ast malloc/free are wrappers on vmalloc ksh93 uses this to delay SIGCLD actions that might call malloc/free
