Brian Raven wrote: > Sorry $Bill, perhaps I am not making myself clear. I was not pointing > out that alarm exists for the purpose of solving the OP's problem, but > simply to counter your statement that it didn't exist.
If exists is the key word here - you are right, but like I said it's not really an alarm. It's a kludge for Win32 that only works in a limited scenario. As far as I'm concerned, it's not alarm but more a time poll function. > Perhaps you should check the man page for alarm. I have just checked the > Solaris 5.8 man page and it clearly says that there is only one timer > per process. It also says that if you call alarm from a bound thread, > using the Solaris threads library, that the signal will be delivered to > that thread. However, it also says that that behaviour is buggy, and is > about to be replaced by the standard (POSIX threads) behaviour of > delivering the signal to the process, which is also the case if you use > the pthreads library. And your point there ? I haven't used Solaris in a while, but it looks like they're going to clean it up. The point is - you can use alarm in a threaded env on UNIX. There are a lot of variables involved, but it's there and it really is an alarm rather than a time poll. There are lots of different thread libraries on lots of different OS's and probably the POSIX approach will unify things. > So, you can only have one timer active, and you can't guarantee which > thread the SIGALRM will be delivered to (unless you have set up your > thread signal masks appropriately), which isn't much use if you want > 'timeout' functionality in multiple threads. Did you catch the part where I said multiplex the alarm signal ? You can catch the alarm in the main thread and signal with another signal to the child thread. I admit it's not totally straight forward, but anyone using threads is probably used to having to work around these kinds of issues. > Sorry $Bill, but just to make it clear, I am just countering your > statement "so if you're on UNIX you'll be fine", which I don't believe > to be true. I have done a fair bit of multithreaded programming on a > number of platforms, and the general consensus seems to be that signals > in a multithreaded environment should be avoided, indeed, some go as far > as to say they are evil. I stand by my stmt that you will be fine on UNIX - as long as you stay within the given functionality of alarm on your platform. Maybe fine is being a bit light about it and a less optimistic word would be more appropriate. But with a little research and some coding, you should be able to implement what he was asking for. The point is that there is a real alarm on UNIX and you can find a way to use it properly in a threaded env. But that's Solaris, he may be on Linux or whatever. So obviously he has to read some man pages to see what he's got available. The poster has probably figured out that it's more trouble than what it's worth by this point in time, but at least we've given him a little insight into alarm. _______________________________________________ ActivePerl mailing list [email protected] To unsubscribe: http://listserv.ActiveState.com/mailman/mysubs
