On Tue, May 14, 2002 at 11:21:23AM -0400, [EMAIL PROTECTED] wrote:
> 
> Jarkko Hietaniemi wrote in reply to Craig Berry:
> 
> !> http://www.openvms.compaq.com/commercial/c/5763p005.html#re_ent
> !
> !It looks like the MULTITHREAD could be it.  How that would require
> !declaring the various _r and _r_proto macros (like d_localtime_r)
> !in configure.com, I dunno.
> 
> I _think_ that Craig was alluding to a possibility that adding
> something like /REENTANCY=MULTITHREAD to ccflags might be a
> way to go (I could be wrong).  I'd like to point out section

Yes, I understood that.  What *I* was alluding to is that
if the MULTITHREAD reveals some _r prototypes in C headers,
the corresponding macros in configure.com need to be tweaked,
too, so that HAS_LOCALTIME_R et al get defined for reentr.h
and reentr.c.

> 1.9.2 of the cited URL document that states that mixing
> AST (asynchronous system traps) type calls with MULTITHREADing
> can be dangerous for a large number of C RTL calls including I/O.
> Note that there are calls to sys$setast() in vms/vms.c and
> they might need some disentangling to avoid the mutex deadlock
> problem alluded to in:
> 
> http://www.openvms.compaq.com/commercial/c/5763p005.html#multithread_restrict_sec
> 
> Peter Prymmer
> 

-- 
$jhi++; # http://www.iki.fi/jhi/
        # There is this special biologist word we use for 'stable'.
        # It is 'dead'. -- Jack Cohen

Reply via email to