Log file Viewer

[03:11:03] <pmxbot> jaraco pushed 5 commits to setuptools (https://bitbucket.org/pypa/setuptools/) :

[03:11:03] <pmxbot> Update to packaging 16.4. Fixes #499.

[03:11:04] <pmxbot> Update changelog

[03:11:04] <pmxbot> Bumped to 20.2.1 in preparation for next release.

[03:11:04] <pmxbot> Added tag 20.2.1 for changeset 1aacb05fbdfe

[03:11:04] <pmxbot> Bumped to 20.2.2 in preparation for next release.

[13:55:57] <pmxbot> jaraco pushed 1 commit to setuptools (https://bitbucket.org/pypa/setuptools/) :

[13:55:57] <pmxbot> Add test capturing expectation that spaces between version specifiers should be allowed. Ref #502.

[17:21:45] <xafer> dstufft: https://github.com/pypa/pip/pull/3512 looks happy :)

[17:28:46] <dstufft> xafer: can you merge master into develop? there's merge conflicts that I'm not 100% sure the resolution to

[17:30:46] <xafer> k

[17:40:08] <xafer> dstufft: done, hopefully correctly :)

[21:16:53] <xafer> Thanks for the release dstufft :)

[22:36:24] <vividboarder> Hello, I have a patch for virtualenv I've submitted on github and I'm wondering if anyone is interested in it or has feedback. https://github.com/pypa/virtualenv/pull/827

[22:41:19] <njs> vividboarder: I believe that virtualenv is mostly unmaintained right now (unfortunately). Also, from quickly skimming that pull request I don't understand why it is needed (I don't know anything about "fish" or "custom path-like variables"), and I can't tell whether it's correct (there's some random sed command for some reason? why? what does it do?), which I'm guessing would make it difficult for anyone reviewing virtualenv patches to quickly decide that

[22:41:19] <njs> it is good to include. (And for most maintainers, if a patch is difficult to review, then it will tend to get pushed off to look at later.)

[22:41:45] <vividboarder> Good call.

[22:42:41] <vividboarder> I'll explain there, but fish is an alternative shell to bash and it uses a character other than : to separate path variables. The sed command replaces that special character with : for compatibility with the subshell

[22:48:01] <tomprince> vividboarder: Perhaps you want https://github.com/adambrenecki/virtualfish ?

[22:50:41] <vividboarder> tomprince oh, that's cool. I'll check it out. The reason that I encounter this bug is that I use fish as my default shell, but I work on repos that use pre-commit. http://pre-commit.com/ And pre-commit creates a bash process and uses virtualenv ¯\_(ツ)_/¯

[22:52:52] <tomprince> vividboarder: Are you also using activate.sh from fish?

[22:53:22] <tomprince> Otherwise, I don't think that variable that you are changing should get set (or, at least it should only be set to something bash will understand).

[22:53:52] <vividboarder> No, I use activate.fish from fish, but then spin up a bash subshell and use activate.sh

[22:54:15] <vividboarder> Or rather, pre-commit spins up a bash subshell and uses activate.sh

[22:54:36] <tomprince> I suspect that it would make more sense to put a fix in activate.fish rather than .sh.

[22:54:40] <vividboarder> Switching that tool to fish doesn't seem feasible since it's not likely installed on many people's devices

[22:55:03] <vividboarder> tomprince maybe.

[22:55:16] <vividboarder> Could always do that and store old paths as bash paths

[22:55:25] <vividboarder> Then in activate.fish when restoring, switch them to fish paths

[22:55:48] <tomprince> Or, can you tell fish to treat that variable like it does $PATH ?

[22:56:29] <vividboarder> Not explicitly. That's also something I could patch, but knowing when to make a variable something it should convert may not be intuitive.

[22:56:53] <vividboarder> Perhaps a flag when setting the variable to indicate that it's a path, even if it's not PATH

[22:57:01] <tomprince> If those scripts are expected to communicate via environment variables, they should be populating them the same, rather than having a hack in one to work around a different one having a weird format.

[22:57:35] <tomprince> I know zsh has something like that, for being able to treat a $PATH like variable as an array.

[22:57:46] <vividboarder> right, but fish and bash don't expect to be communicating.

[22:57:57] <vividboarder> Or at least, bash doesn't expect to get a variable in fish format

[22:59:32] <tomprince> Well, you are expecting them to communicate.

[23:02:16] <vividboarder> Yea

[23:02:20] <vividboarder> That's true....

[23:03:21] <njs> I'm confused about how the fish thing even works, given that PATH is something that's interpreted by libc

[23:05:34] <vividboarder> I wonder if that's why PATH is "converted" when you have a bash subshell because it's just showing the actual path.

[23:05:49] <vividboarder> But for any other variable, like OLD_PATH, it's being stored as-as

[23:05:53] <vividboarder> as-is*

[23:07:46] <vividboarder> I think this could be done all in activate.fish... I'll give that a shot.

[23:08:24] <vividboarder> I think I can store my OLD_PATH as (bash -c 'echo $PATH') to get it to store the OLD_BASH_PATH

[23:09:29] <tomprince> I guess https://github.com/fish-shell/fish-shell/issues/436 is probably relevant.

[23:16:21] <tomprince> vividboarder|afk: You shouldn't need to use bash to get that.

[23:42:42] <vividboarder> I could use that same sed command I guess