Darwin and SunOS support

[Maciek416]

This PR adds support for killing process trees in Darwin (i.e. Mac OS X) and SunOS (i.e. Joyent SmartOS, etc).

Notes:

• Because Darwin and SmartOS lack a --ppid PID option in the ps command, for those two we simply get a formatted list of processes in ppid, pid format and later filter through them with split() and map() and filter().
• Darwin and SmartOS also lack a --no-header option, so we skip the header the above split/map/filter implementation.
• For Linux, instead of working through the string character by character, we similarly use split() and map().
• For all Unix OSes, we call cb() when we've run out of pids to process, not necessarily when we've got an exit code of 0 from ps. This is because in Darwin and SmartOS, we're running ps in a mode that doesn't return a non-zero code, ever.
• For all Unix OSes, we only check the ps exit code to check if it's okay to parse ps output. Recursion into deeper calls to buildProcessTree() are determined by whether we've found any pids.

[billiegoose]

Awesome... I definitely would like tree-kill to support Darwin and SunOS. I would prefer to do it without adding a bunch of conditional code based on OS-checking though. :-/ I'll look closer to see if this can be simplified without sacrificing support for all the OSes.

[billiegoose]

Here's what I'm thinking. Rather than insert OS-specific conditionals scatter shot throughout the code, let's keep all that we can in the top most function where we check for windows. Instead of the if/else we can use a switch statement on process.platform. We can keep the original buildProcessTree as it is and make a new one for the Darwin/SunOS version of ps.

I really don't want to break things unintentionally, and by keeping the different OSes siloed from each other, people can safely edit the version for the OS they use and be confident they didn't break the module for other OSes they can't test on.Does any of that make sense? I'll make a commit later today building out the structure I'm talking about. Time to start a 'develop' branch...

[billiegoose]

@Maciek416, take a look at 9331318 and see if that makes sense. Any chance you could fill in buildProcessTreeDarwin() and buildProcessTreeSunOS() and test it on those platforms for me?

[Maciek416]

Hey @wmhilton, I'll try to take a look this week. Thanks for following up !

[billiegoose]

@Maciek416 any update?

[UltCombo]

Can't we use pgrep -P [PPID] in Darwin? From the man page:

DESCRIPTION The pgrep command searches the process table on the running system and prints the process IDs of all processes that match the criteria given on the command line.

The following options are available: [...] -P ppid Restrict matches to processes with a parent process ID in the comma-separated list ppid.

That is, pgrep -P parentPid will return a new-line separated list of child processes. Couldn't it be used to build the process tree?

[UltCombo]

In fact, I believe pgrep -P [PPID] should also work on Linux distros. Is there any reason to use ps instead of pgrep?

[billiegoose]

Is there any reason to use ps instead of pgrep ?

Well, ps is part of the POSIX standard. I don't think pgrep is. Can we be certain that pgrep is installed on every system? It could be a breaking change for some people... I'd rather not risk that.

That bring said, if it's already broken on OS/X, I'm fine with fixing it as long as we don't touch the Linux code.

[billiegoose]

Anybody want to insert some Darwin code in my lovely case statement? https://github.com/pkrumins/node-tree-kill/tree/develop?files=1

[UltCombo]

Well, ps is part of the POSIX standard. I don't think pgrep is. Can we be certain that pgrep is installed on every system? It could be a breaking change for some people... I'd rather not risk that.

Oh, I see. Makes sense.

Anybody want to insert some Darwin code in my lovely case statement?

I believe I should be able to submit a patch in 10 hours or so.

[UltCombo]

10