rename tests of packaging and improve them to avoid spurious system-dependent test failures

[zooko]

This is a ticket for me to manage small clean-up patches which I would like to see applied but not before 1.8.2.

[zooko]

please review: attachment:tighter-catch.darcs.patch​

[zooko]

please review: attachment:improved-package-tests.darcs.patch​

[davidsarah]

scripts/common.py: don't assume that the default alias is always 'tahoe' (it is, but the API of get_alias doesn't say so). refs #1342

[davidsarah]

+1 on attachment:tighter-catch.darcs.patch​

[davidsarah]

Reviewing attachment:improved-package-tests.darcs.patch​ :

test-dont-install-newer-dep-when-you-already-have-sufficiently-new-one.py is way too long a filename. Also the test is about which dist is built and tested (by setup.py test) when an existing one is installed, not about whether the new dist is installed. (It should be "dist", not "dep", since we're talking about a particular version/build.)

How about test-be-satisfied-with-new-enough-dist.py?

test-dont-use-too-old-dep.py (which should be test-dont-use-too-old-dist.py) has this:

# The goal is to turn red if the build system tries to use the
# source dist when it could have used the binary dist.

which I think is a stale comment from a different test.

[davidsarah]

docs/running.html: reference quickstart.html instead of install.html, and don't refer to 'installing' Tahoe-LAFS (which is not what the quickstart instructions say to do). refs #1342

[davidsarah]

Given that #1355 is a potential regression, we probably shouldn't apply these quite yet.

[davidsarah]

relnotes.txt: forseeable -> foreseeable. Don't claim to work on Cygwin (which has been untested for some time).

[zooko]

please review: attachment:docfix.darcs.patch​

[ChosenOne]

+1 on attachment:relnotes.darcs.patch​,

attachment:docfix.darcs.patch​ and attachment:tighter-catch.darcs.patch​

Please review attachment:improved-package-tests.darcs.patch​ - I didn't fully grasp that one ;)

[david-sarah@…]

In 16a2f71eea022c92:

relnotes.txt: forseeable -> foreseeable. refs #1342

[david-sarah@…]

In 2dd742b24808d54a:

relnotes.txt: don't claim to work on Cygwin (which has been untested for some time). refs #1342

[zooko]

• applied attachment:tighter-catch.darcs.patch​ in b30a269ec2477e84
• applied attachment:docfix.darcs.patch​ in 9619812a9db74aab
• applied attachment:relnotes.darcs.patch​ in 2dd742b24808d54a and 16a2f71eea022c92
• added a M-x whitespace-cleanup in f251bbece5f8f188

[davidsarah]

Replying to ChosenOne:

Please review attachment:improved-package-tests.darcs.patch​ - I didn't fully grasp that one ;)

I saw the test-with-fake-dists.py test, that this patch is based on, looking for a package "fakedependency" on PyPI. Maybe we should split this out into a separate ticket and think more carefully about how to do it.

[davidsarah]

Review still needed for default-alias-cleanup.darcs.patch​.

[zooko]

+1 on attachment:default-alias-cleanup.darcs.patch​

[zooko]

Oh look: ChosenOne went back and edited his earlier comment to add +1's for attachment:default-alias-cleanup.darcs.patch​ and attachment:docs-running-no-install.darcs.patch​ . :-)

ChosenOne: next time please add a new comment instead of editing your old one so that people who are just checking new comments will notice. :-)

[davidsarah]

docs-running-no-install.darcs.patch​ is obsoleted by 299e8ad5795d3c22 (which deletes docs/running.html and changes docs/running.rst​ to point to docs/quickstart.rst).

[david-sarah@…]

In 49fd2e6e564a0e01:

scripts/common.py: don't assume that the default alias is always 'tahoe' (it is, but the API of get_alias doesn't say so). refs #1342

[zooko]

If I understand correctly, the only patch that hasn't been reviewed and committed or else superceded is attachment:improved-package-tests.darcs.patch​. Changing the name of this ticket!

[Brian Warner <warner@…>]

In d86776560411accc:

modify build_helpers files

Should close #1342. This makes the actual changes to the two test
files (separated from the 'rename' patch to avoid VC complications).

[Brian Warner <warner@…>]

In d86776560411accc:

modify build_helpers files

Should close #1342. This makes the actual changes to the two test
files (separated from the 'rename' patch to avoid VC complications).

[davidsarah]

The original improved-package-tests.darcs.patch​ renamed some files, and that change hasn't been applied. However, I think the new names (e.g. misc/build_helpers/test-dont-install-newer-dep-when-you-already-have-sufficiently-new-one.py) were excessively long.

Also I don't really understand what improvement this patch is trying to make. AFAICS it hasn't been reviewed, and I was waiting for Zooko to answer comment:12.

[warner]

Actually I landed a patch just before d86776560411accc (probably 533e4bc813d299e7) that did the renames, but I forgot to include a "refs #1342", so it didn't get mentioned here. (I split the patch into two pieces, a file-move followed by file-modify, because I had to land it via git (since darcs went into an O(N!) death spiral trying to apply the dpatch) and I wanted to avoid problems in the git/darcs bridge).

I agree that the new names are excessively long, but they're in a fairly obscure corner of the tree so I decided I'd let it go. I skimmed over the changes and they seemed ok. But I did miss that you were asking zooko for something in comment:12 .. sorry about that!

In the buildbot run just now, it didn't look like it was trying to get "fakedependency" from pypi, but I might be missing something. Could you take a look at that run and see if you're still seeing the problem you observed earlier?

[Brian Warner <warner@…>]

In 488b6f8ccdd50eb3:

test-dont-use-too-old-dep.py: fix tarfile timestamps

It turns out that TarFile?.addfile() doesn't provide a reasonable default
timestamp, resulting in files dated to 1970 (they're probably wearing
bell-bottoms and listening to disco too). Then, when the bdist_egg command
tries to create a *zip*file with those files, it explodes because zipfiles
cannot handle timestamps before 1980 (it prefers boomboxes and jackets with
straps on the shoulders, thank you very much).

This puts a modern time.time() on the members of the tarfile, allowing future
cryptocoderarchaeologists the opportunity to make fun of fashion trends from
the user's chosen era, rather than an artificially older one.

refs #1342

[Brian Warner <warner@…>]

In 488b6f8ccdd50eb3:

test-dont-use-too-old-dep.py: fix tarfile timestamps

It turns out that TarFile?.addfile() doesn't provide a reasonable default
timestamp, resulting in files dated to 1970 (they're probably wearing
bell-bottoms and listening to disco too). Then, when the bdist_egg command
tries to create a *zip*file with those files, it explodes because zipfiles
cannot handle timestamps before 1980 (it prefers boomboxes and jackets with
straps on the shoulders, thank you very much).

This puts a modern time.time() on the members of the tarfile, allowing future
cryptocoderarchaeologists the opportunity to make fun of fashion trends from
the user's chosen era, rather than an artificially older one.

refs #1342

[davidsarah]

Replying to warner:

In the buildbot run just now, it didn't look like it was trying to get "fakedependency" from pypi, but I might be missing something. Could you take a look at that run and see if you're still seeing the problem you observed earlier?

Will do.