-
-
Notifications
You must be signed in to change notification settings - Fork 11.4k
Conversation
@UniqMartin FYI, it looks like the travis test failed due to |
Is there any way to disable the self-update feature? That sort of thing doesn't really play nice with package management:
|
Fair point. AFAICT, no there's no way to do that, without explicitly removing the upgrade functionality. The self upgrading was introduced in v0.5 by commit mitsuhiko/pipsi@7442f7c and looking at lines 156 -- 181 it doesn't look like there's much that can be done about this, short of a patch that would disable upgrading anything and everything installed by pipsi. |
@zbeekman That seems to be happening more often in recent times, unfortunately. But as long as the default build (on our Jenkins CI) is fine, you can ignore the Travis CI failure. I agree that it is annoying though and gives a false impression in the PR list. |
It's not necessarily a blocker but ideally we'd file an Issue upstream asking for a way to disable the self-update mechanism impacting the tool itself specifically. |
I'll go ahead and do that now: mitsuhiko/pipsi#68 |
This probably needs to resource |
You should be able to install something as a test; $HOME is set to a temporary path during install and test so pipsi shouldn't actually affect a user's system. |
@tdsmith Thanks for the insight! I'll go ahead and resource pip, and try a test install... I guess pipsi can uninstall software too... so we just need to find a suitable package with minimal dependencies to install and uninstall in |
@tdsmith I have resourced How are things looking now? |
Mysterious test failure! |
I'm flummoxed |
My best guess is that this has to do with the environment? Maybe I need I'd appreciate any help or insight @tdsmith or @RonnyPfannschmidt or @mitsuhiko can provide. |
You need the pipsi bin dir and virtualenv folders within the sandbox |
@tdsmith Do you know how to accomplish this? Should I add a |
I think both of those things should be relative to $HOME so they already are within the sandbox. |
Local readout is:
Structure is more or less:
Passes alright though, no strange failures locally. |
Thanks @DomT4, yes that's what I saw locally too, hence why I pushed it up to the PR. @RonnyPfannschmidt Any more detailed thoughts? Have you taken a look at the output on Jenkins? |
I doubt it's related but you've added a |
Thanks for pointing that out. On Sun, Feb 21, 2016 at 10:31 AM Dominyk Tiller notifications@github.com
|
Trying again because I appear to have forgotten to push after resourcing |
Looks like it, yeah 😕. |
Well, if the folks over at the pipsi project don't weigh in soon we may need to close this PR; I'm pretty stuck. |
It's been over a month. Closing this out as infeasible. If you hear back from upstream or other things change, feel free to resubmit this at the new homebrew/core repo. |
pip script installer: install Python scripts with pip to their own
virtualenv to prevent cross contamination and to sandbox them
A note on the
test do
block. Currently,pipsi
provides 4 commands:install
list
uninstall
upgrade
The only one of these that seems suitable for a
test do
block is potentially thelist
command, although even this will be potentially uninformative. Any other command will have undesired and un-requested effects on the user's system. Unfortunately, for the time being, thelist
command crashes when nothing has been installed yet. I made the upstream aware of this in mitsuhiko/pipsi#66 so, until this issue is resolved, the only tests I can seem to think of are simple--help
or--version
tests. Even whenlist
comes online when nothing is installed, this doesn't seem to be more rigorous than--version
or--help
.