"Could not find builtins" error

[wizzardx]

Hi there.

With the latest dev version of mypy (currently 47b40a7) - since my most recent git pull, I get this error when trying to run mypy:

---

test.py:1: error: Could not find builtins

Where this is the complete contents of test.py:

---

!/usr/bin/env python3.4

print("HELLO WORLD")

And the command-line is:

---

mypy test.py

(Where I am under a Python virtualenv, where the latest dev version of mypy has been installed).

I ran a git bisect to track down where this issue started, and ended up at this commit where - in my testing - this problem starts manifesting:

---

commit 2c009ca
Author: Ben Longbons b.r.longbons@gmail.com
Date: Tue Jun 30 22:18:20 2015 -0700

Move script to mypy.main module and just leave stubs

---

This is a breaking issue for me - for now I need to do my dev/testing against the commit prior to that, 2e591dc, where the problem has not yet manifested.

Please advise?

[o11c]

Suspect that something is wrong with setup.py, but travis doesn't test that yet.

[o11c]

Or it might be a problem with default_data_dir etc ... those functions are definitely suspicious.

[o11c]

Have reproduced after modifying the way the tests are run.

WIP: https://github.com/o11c/mypy/tree/setup

[o11c]

@JukkaL The PR I'm about to create will be high priority.

[t2y]

I had same issue. Is there any workaround?

[o11c]

@t2y Workaround is to set PYTHONPATH (and PATH to scripts subdir) to the source dir of mypy instead of installing it.

But I think a fix is ready on my fork, use:

pip install git+https://github.com/JukkaL/o11c@setup

[o11c]

My fix has been rejected, just use my fork instead permanently.

[JukkaL]

@o11c: I did not reject your fix, I asked you to resend it with unrelated changes removed and with an explanation of how the fix works. You did not wish to do this but started complaining publicly, so I had to fix it myself. Emergency fixes should be short and to the point. Time and time again a rushed fix has introduced additional problems (see http://www.infoworld.com/article/2608134/microsoft-windows/the-new-kb-2919355-windows-8-1-update-causes-more-problems-than-it-fixes.html for example).

@wizzardx @t2y: Please let me know if it works for you now.

[JukkaL]

Postmortem:

The new test runner moved most of the logic from scripts/mypy to scripts/main.py. However, some of the logic in the file assumed that the file was living in the original path, and moving the file broke running mypy when installed. I fixed it by passing the path of script in the new stub scripts/mypy to scripts/main.py. This change also uncovered another bug that I also fixed. I'm not sure whether the latter bug actually caused any problems until this fix.

[wizzardx]

Already did some testing on my end, but didn't want to fan a possible flame thread :-). Good damage control there.

Both your updates and o11c's fork fixed the problem I reported earlier, thanks.