[vwnc] ExternalProcess on Mac OS X

Thomas Cleenewerck wrote:
> Michael,
>
> Circumventing the path problem, I now use the following working
> example code:
>
> ExternalProcess fork: '/usr/local/bin/jhead' arguments: #('-autorot
> /Users/thomas/Pictures/2010/DSC_8249.JPG')
>
> However jhead uses jpegtran, another unix program, which fails to load
> because jpegtran could not be found.  I think I'm experiencing the
> same path problem. This is the error I get.
>
> sh: jpegtran: command not found
> Error : Problem executing specified command
> in file '/Users/thomas/Pictures/2010/DSC_8249.JPG'
>
> Is there a way to set the start path of a process?
>
Pretty much - no. Current path is actually a property of terminals,

so
if you wanted to do that, you could fire up the shell, cd to the
directory you want, then run your command, eg:

ExternalProcess execute: '/bin/sh' do: [:in :out |
  out nextPutAll: 'cd ', path; cr.
  out nextPutAll: 'theprogram; exit'; cr.
  in upToEnd]
     errorStreamDo: [:err | err upToEnd]

The alternative (and less desirably) is to change the current directory
in VW before running the command, then changing back as soon as the
command is launched.

PATH is a property of terminals too.. you could read in the PATH
variable and use it to find the location of the executable.

Cheers,
Michael
> Thomas.
>
> p.s. in my terminal this works fine.
>
> On 27 Jan 2009, at 20:26, Michael Lucas-Smith wrote:
>
>> Thomas Cleenewerck wrote:
>>> Hello,
>>>
>>> I have used ExternalProcess before and it worked perfectly. However
>>> when I used it  with unix program jhead it failed and gave the
>>> following criptic error: 'unhandled exception: Arguments invalid --
>>> wrong class'.
>>>
>>> "failing code"
>>> proc := ExternalProcess new.
>>> proc  execute: 'jhead'
>>>           arguments: rotarguments
>>>       do: [:in :out | nil ] errorStreamDo: [:err | Transcript show:
>>> err  contents ].
>>> proc release.
>>>
>>> When I tried the full path of the unix command it worked, although
>>> the  full path is a part of  my path environment variable. Any
>>> suggestions  what to do to resolve this?
>>>
>>> "workaround"
>>> proc := ExternalProcess new.
>>> proc  execute: '/usr/local/bin/jhead'
>>>           arguments: rotarguments
>>>       do: [:in :out | nil ] errorStreamDo: [:err | Transcript show:
>>> err  contents ].
>>> proc release.
>>>
>>>
>> I'm not sure what your rotarguments are, should be an array.
>> Also, you need to read from the :in variable otherwise, the process
>> will end immediately. You need to attempt to read from both in and error
>>
>> Cheers,
>> Michael
>
>

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc

>     Pretty much - no. Current path is actually a property of terminals,
>
>
> More correctly it is a property of shells, but for utility it is also
> built into the C library.  See e.g. execlp(3), execvp(3) & execvP(3).

These remarks surprise me.

I always thought that 'the environment' was a property of processes.

If a process spawns a new one it can set up the env for the new process
or leave it implied (in which case the child gets a copy of the parent
process' env).

So I /guess/ you can simply set the path in the environment of the VW
process, and the spawned utilities will inherit it.

Something like:

  CEnvironment setenv:
    (CEnvironment getenv: 'PATH') , ';/usr/local/bin'

Another option would be to launch a shell in 'login shell' mode (so it
sets the users preferred path), but I don't know how that exactly works
and expect it to be very brittle (permission problems depending on how
apps are launched and such).

HTH,

Reinout
-------

_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc