On Wed, Jun 19, 2019 at 12:04 PM Stephen Smalley <sds@xxxxxxxxxxxxx> wrote: > On 6/19/19 10:40 AM, Paul Moore wrote: > > On Wed, Jun 19, 2019 at 10:26 AM Stephen Smalley <sds@xxxxxxxxxxxxx> wrote: > >> > >> /usr/bin/python is no longer installed by default, so this test > >> script fails on default installs. > >> Best practice guidance appears to be to switch to using > >> /usr/bin/python3 instead, per > >> https://docs.fedoraproject.org/en-US/packaging-guidelines/Python/ > >> > >> Fixes: #46 > >> Signed-off-by: Stephen Smalley <sds@xxxxxxxxxxxxx> > >> --- > >> tests/overlay/access | 2 +- > >> 1 file changed, 1 insertion(+), 1 deletion(-) > > > > I think it would be better to reference the Python upstream guidance > > regarding "python3" instead of the Fedora packaging guidelines. > > Keeping things distro agnostic is a good goal whenever possible. > > > > * https://www.python.org/dev/peps/pep-0394 > > When I looked for guidance in response to the github issue for > selinux-testsuite [1], there seemed to be some ambiguity left by pep-0394: > > 1) It says "One exception to this is scripts that are deliberately > written to be source compatible with both Python 2.x and 3.x. Such > scripts may continue to use python on their shebang line.". This leaves > open the possibility of staying with "/usr/bin/python" as far as the pep > is concerned since the script itself is not version-specific in any way, > but doing so is not permitted by the Fedora guidelines and breaks on > default installs of Fedora. [SIDE NOTE: for some reason my mail is acting a bit odd - this got buried in a folder and didn't bubble up like normal - my apologies on the delay, I didn't see it until I was checking for patches today.] Yes, I saw that too but considering that we are about six months away from python2 going EOL I think it is reasonable to ignore python2 for upstream work at this point. > 2) It doesn't specify whether or not one should use "/usr/bin/env > python3" or just "/usr/bin/python3". The github issue reported had > originally suggested the former, but this is also prohibited by the > Fedora guidelines. My initial reaction was that we should "/usr/bin/env" until I saw that bit in the Fedora guidelines, which I honestly don't understand. Does anyone with a better understanding of cross-distro userspace know why "/usr/bin/env" is bad? > That's why I went with the Fedora guidelines. Happy to use whatever but > at least that particular PEP doesn't clearly motivate the use of > "/usr/bin/python3" here. > > [1] https://github.com/SELinuxProject/selinux-testsuite/issues/46 Fair enough, I'll go ahead and merge it since it does fix a real problem. I did make once change to convert the "Fixes: #46" to use the full URL, e.g. "Fixes: https://github.com...";; I think this helps provide some necessary context (what if 10yrs from now we aren't on github, what does "#46" reference?). This is starting to stretch the topic a bit, but as far as "use whatever" is concerned, I'd really like to see us keep moving towards a more distro agnostic approach when it comes to upstream development/testing/etc. While SELinux has always had a "special" relationship with Fedora/RH, we've seen SELinux grow quite a bit over the years and I think it would be good to start trying to bring other distros up to the same level as Fedora when it comes to SELinux support. I think one easy first step is to start looking beyond the Fedora guidelines and see what other distros do/recommend. > >> diff --git a/tests/overlay/access b/tests/overlay/access > >> index 5a47eb145904..e50d5b402ebd 100755 > >> --- a/tests/overlay/access > >> +++ b/tests/overlay/access > >> @@ -1,4 +1,4 @@ > >> -#!/bin/python > >> +#!/usr/bin/python3 > >> import os, sys > >> try: > >> acc=eval("os.%s" % sys.argv[2]) > >> -- > >> 2.20.1 > > > -- paul moore www.paul-moore.com
