Hi Armin and Maciej, First, thanks for your mails.
> The fact that the directory name is predictable is very useful in various
> situations, so it's unlikely that we can be convinced to change that.
Hm, let's try and understand a little bit more about these temporary
directories.. By "predictable", who has to predict them (from PyPy's point of
view)?
If it's limited to a single process, then would it not be possible create a
temporary directory in which PyPy places all the other directories? This
seems the same as just doing:
import tempfile
os.environ['TMPDIR'] = tempfile.mkdtemp()
[...]
.. or similar. (I seem to recall discussing something along these lines on
IRC, but I can't remember the outcome)
The other issue seems to be where these directories are used: Can I assume
that--as the code is part of the pylib--that other tools such as py.test use
these directories and thus a "general" solution needs to be found? (If they
were limited to the build system then I could probably hack around it using
a shell variant of the above.)
Anyway, whilst your SHA1 dot-file proposal sounds interesting and promising,
I wonder if we're missing something obvious.. Hope you can help.
Regards,
--
Chris Lamb, UK [EMAIL PROTECTED]
GPG: 0x634F9A20
signature.asc
Description: PGP signature
