[9fans] Threadcreate() failure value?

[9fans] Threadcreate() failure value?

[ISHWAR RATTAN]

The man page does not mention what is the value returned
by threadcreate() and proccreate() on failure? Can one safely
assume that it is -1?

-ishwar

Re: [9fans] Threadcreate() failure value?

[Russ Cox]

> The man page does not mention what is the value returned
> by threadcreate() and proccreate() on failure? Can one safely
> assume that it is -1?

thread(2):

        Thread library functions do not return on failure;
        if errors occur, the entire program is aborted.

Re: [9fans] Threadcreate() failure value?

[ISHWAR RATTAN]

On Wed, 30 Nov 2005, Russ Cox wrote:
> thread(2):
>
>        Thread library functions do not return on failure;
>        if errors occur, the entire program is aborted.

Interesting. If threads are being created by a min thread on the
fly to handle some task, it means all threads (including main thread)
disappear?? What if there is a need for active threads to complete
the assigned tasks?

-ishwar

Re: [9fans] Threadcreate() failure value?

[Russ Cox]

> On Wed, 30 Nov 2005, Russ Cox wrote:
> > thread(2):
> >
> >        Thread library functions do not return on failure;
> >        if errors occur, the entire program is aborted.
>
> Interesting. If threads are being created by a main thread on the
> fly to handle some task, it means all threads (including main thread)
> disappear??

The only way these can fail is if you run out of memory
or you run out of processes.

> What if there is a need for active threads to complete
> the assigned tasks?

Then you should fix the memory or process leaks.

Russ

Re: [9fans] Threadcreate() failure value?

[Ronald G Minnich]

Russ Cox wrote:

>>What if there is a need for active threads to complete
>>the assigned tasks?
> 
> 
> Then you should fix the memory or process leaks.
> 
> Russ


what if your create failed due to some other process having the process 
leak, and you'd like your process, which has done nothing wrong, to get 
out of the mess cleanly?

ron

Re: [9fans] Threadcreate() failure value?

[Russ Cox]

> what if your create failed due to some other process having the process
> leak, and you'd like your process, which has done nothing wrong, to get
> out of the mess cleanly?

Then you will need to change the thread library and then
edit all the programs that use it to check the return values
on threadcreate and proccreate.

Russ

Re: [9fans] Threadcreate() failure value?

[Ronald G Minnich]

Russ Cox wrote:
>>what if your create failed due to some other process having the process
>>leak, and you'd like your process, which has done nothing wrong, to get
>>out of the mess cleanly?
> 
> 
> Then you will need to change the thread library and then
> edit all the programs that use it to check the return values
> on threadcreate and proccreate.

or ....

provide versions of the functions that return on error, for those cases 
where they are needed. And then have the "normal" function call this 
function, and on failure, act as it acts now (i.e. exit)?

I realize that just blowing the process away on this type of failure is 
a very reasonable thing to do, but in a very few cases here, we need to 
recover and stumble along in the case of such failure, and try to at 
least get to a sane state before dying.

ron

Re: [9fans] Threadcreate() failure value?

[Russ Cox]

> I realize that just blowing the process away on this type of failure is
> a very reasonable thing to do, but in a very few cases here, we need to
> recover and stumble along in the case of such failure, and try to at
> least get to a sane state before dying.

If you really have a need for this, then let me know and I'll do something.
The most likely solution is to have a user-supplied error function
like the regular expression and draw libraries have.

But I'm not going to spend time on this for its own sake.

Russ

Re: [9fans] Threadcreate() failure value?

[Martin Harriss]

>>what if your create failed due to some other process having the process
>>leak, and you'd like your process, which has done nothing wrong, to get
>>out of the mess cleanly?

One of the problems is that the program that finds itself unable to 
proceed for lack of resources may not be able to recover because it 
needs resources to run the recovery mechanism.

This was dicussed a long time ago wrt memory allocation for Modula 3. 
See, for example:

http://www.elegosoft.com/pm3/intro/questions/new.html

(I'm sure that somewhere I've seen a more basic description of why 
handling out-of-memory conditions is diffucult, but I can't find it 
right now.)

I'd be the first to admit that this is a somewhat different case because 
M3 relies heavily on dynamic memory allocation and garbage collection, 
but I think that in general terms the issues are the same.

Martin