[Twisted-Python] Twisted Trial changes working directory

[Gerrat Rickert]

>-----Original Message-----
>From: twisted-python-bounces at twistedmatrix.com [mailto:twisted-python-
>bounces at twistedmatrix.com] On Behalf Of Jean-Paul Calderone
>Sent: Friday, July 03, 2009 11:27 AM
>To: Twisted general discussion
>Subject: Re: [Twisted-Python] Twisted Trial changes working directory
>
>On Fri, 3 Jul 2009 10:17:12 -0400, Gerrat Rickert
><grickert at coldstorage.com> wrote:
>>I have a (Twisted) program that saves & uses a file in the current
>>working directory.
>>When running tests for the program, Trial changes the working
>directory,
>>causing this file to be written into the '_trial_temp' directory
>>instead.
>>...so the program works differently when run via Trial vs. in normal
>>use.
>
>The only difference I can infer from your description is whether the
>file
>in question is created in the directory you invoke the program from (in
>the non-trial case) or in the _trial_temp subdirectory of that
directory
>(in the trial case).  It's not obvious to me why this might be a
>problem.

...and it turns out not to really be a Trial-specific problem - more a
problem with my expectations.
The file takes a while to generate, and I just expected it to be there
after running the program the first time.
- it will get regenerated if it doesn't exist, and I can easily copy it
to the _trial_temp directory for testing

>
>>
>>There are a number of solutions to this, but I'd just like to
>understand
>>the reasoning behind this surprising behaviour.
>>The docs say that this: "...allows them to write whatever data they
>like
>>to disk, and not worry about polluting the current working directory"
>>...couldn't they just do this without changing the working directory?
>
>This is an ancient feature of trial, from the days when Twisted's own
>test
>suite (the primary driver of trial development for many years) was even
>more crufty than it is today.  In those days, many tests just created
>files
>in the working directory.  The _trial_temp feature provided a single
fix
>which addressed the mess created by all of the tests in one fell swoop.
>The
>behavior is still convenient for tests which aren't written carefully
in
>this regard.
>
>Jean-Paul

Thanks for the explanation.