On 27/08/2011 10:38 AM, Enrico Forestieri wrote:
On Sat, Aug 27, 2011 at 02:07:02AM +0200, Julien Rioux wrote:
Due to my playing with git, the commits r39525 and 39526 happened
simultaneously, while I only meant to commit the first.
My question is: What problem is addressing r39526?
subprocess.Popen itself requires 2.4, only the windows script
already requires this python version. Is it time to bump the
requirement for everyone?
I am against bumping requirements without any good reason.
One issue is that os.popen relies on the shell to launch external
scripts, with for example issue #7718 as consequence. I think a patch
using subprocess could fix this outstanding issue. The commit r39526 was
only a first step, simply replacing os.popen by current best-usage
suggestions from the python docs. Since subprocess.check_output is
definitely too new to justify the version bump, I have reverted to using
only subprocess.Popen in r39533. Next step would also remove the
shell=True part, so that nothing depends on the shell anymore.
I will keep working locally until the whole thing is figured out.
At the moment I would either revert the whole thing, or if it's ok we
wait a bit until it is proven that there is real benefit, and then go
ahead and update the README.
Cheers,
Julien
