SUnit improvements need review and feedback

Hola,

for the record, I'm using OSSubprocess in several projects and it's pretty robust.

The problems I had were because either:

- I was using it wrongly. e.g., I was deadlocking my process due to a process that was writing a lot to stdout. I fixed this after I read the entire documentation. I think that there should be a better warning at the beginning of the docs because all the starting examples do not use polling, and then this may create problems (because people read the documentation until they found what they needed, they will not read the entire thing ^^)

- Or there was a library that was creating some noise with it. I remember at some point Seamless and OSSubprocess together were a crash factory. So I'm restarting now again with those two.

The thing is that it is really frustrating to have crashes all day long with something that should not crash :( It's not that I'm messing with the VM, or that I do not know what I am doing...

-------- Original Message --------

Guille, 

Just for the record, there has been some crashes in Pharo using OSSubprocess, but I (we) were never able to reproduce nor found the cause, and even less a fix :(

So....any help in this direction is really appreciated as I want OSSubprocess to be robust. 

Cheers, 

On Mon, Sep 5, 2016 at 2:53 PM, Guillermo Polito <[hidden email]> wrote:

I mean crash.

Somehow my image's processes were killed and I only had the UI process: no delay process, no finalization, no idle process. Then, when the VM tries to suspend the current process and change to a second one, it finds none and it just quits.

So it's not a VM problem I think but mostly something in the image.

And in any case, now that I'm with a colder head (because I was mad after losing a couple of hours of work this afternoon :)) ):

 - first, I could not load OSSubprocess in latest pharo. I saw there were two test watcher processes and *killed* them.

 - then, I could load OSSubprocess, but it started crashing the image when running the tests from nautilus.

 - I tested with complex and simple OSSubprocess calls and I got always the same result

 - I even removed the child watcher process to see if that was the problem, but still crashing.

In any case, I'm not sure that this is because of this addition, it's just a hunch for the moment.

On Mon, Sep 5, 2016 at 7:47 PM, stepharo <[hidden email]> wrote:

ARgh, since this got integrated I keep crashing images...

do you mean real crash or DNU.
Because indeed I got some crashes too.

--

Mariano
http://marianopeck.wordpress.com

ARgh, since this got integrated I keep crashing images...

I'm just executing some OSSubprocess like this:

process := OSSUnixSubprocess new

command: 'ls';

redirectStdout;

runAndWaitOnExitDo: [ :command :outString | ^ outString ].

and my image crashes with a:

(scheduler could not find a runnable process)

So this is messing up badly the scheduler. We have to fix this :/

Tx denis.

I wonder what is the reason of the crashes I experienced.