Log file Viewer

[21:23:20] <lifeless> hihi

[21:23:27] <lifeless> dstufft: o/

[21:29:28] <dstufft> lifeless: heya

[21:36:59] <lifeless> trying to figure this failure out

[21:37:10] <lifeless> it would be nice if much more state was logged/captured

[21:42:46] <lifeless> ok, the source tree is corrupt

[21:43:23] <lifeless> it has a PKG-INFO in it, for instance

[21:47:43] <lifeless> ahh

[21:47:49] <lifeless> there's no delete marker in that dir

[21:49:07] <xafer> is it a vcs requirement ?

[21:50:07] <lifeless> no, Upper, so tarball

[22:48:16] <lifeless> dstufft: is it possible the existing only_download use in unpack_url is inverted?

[22:48:29] <lifeless> dstufft: erm, probably from req_set.py

[22:48:55] <lifeless> dstufft: My hypothesis is that we haven't noticed because the build dir containing all the locations is itself deleted

[23:01:55] <lifeless> ah yes

[23:02:00] <lifeless> now I think I know :)

[23:04:08] <lifeless> dstufft: the use of --download-dir to imply --no-install is odd, because no-install implies unpacking-and-keeping, but download just implies keeping.

[23:13:09] <lifeless> lets see what fallout this creates

[23:34:46] <lifeless> dstufft: I invoke thee :)

[23:36:47] <lifeless> dstufft: I think test_download_should_not_delete_existing_build_dir is wrong

[23:40:21] <lifeless> dstufft: but perhaps I misunderstand it. What is the intent ?

[23:40:24] <lifeless> qwcode: ^

[23:41:28] <lifeless> for context, I've discovered that we don't keep unpacked dirs correctly- we were keying off of having a download dir, but -d shouldn't imply keeping the unpacked source

[23:41:35] <lifeless> --no-install was for that

[23:41:58] <qwcode> lifeless, hmm, I recall documenting this I think in an issue maybe.. looking....

[23:42:11] <lifeless> I split this outinternally into a keep_extracted flag for req_set

[23:42:25] <lifeless> the only testthat failed was test_download_should_not_delete_existing_build_dir

[23:42:34] <lifeless> and its failure is that we see more source in the build dir

[23:42:42] <lifeless> which AFAICT from the --no-install docs is *correct*

[23:43:04] <lifeless> thus me speculating that the test is wrong - if it captured behaviour from after --no-install was deprecated

[23:48:00] <qwcode> not sure that the test is wrong per se, but probably ok, to change that behavior I think.

[23:48:27] <qwcode> can't find the issue I was thinking of

[23:50:37] <qwcode> hold on... looking at test again...

[23:53:20] <qwcode> lifeless, why would you want "pip install --no-install" to clean out the pre-existing build dir?

[23:54:00] <lifeless> qwcode: I don't, and it doesn't

[23:54:07] <lifeless> sec

[23:55:14] <lifeless> qwcode: https://github.com/rbtcollins/pip/commit/5f076e21e5a3ff6b1a460c244c8786f3ad0f3ef3