desert-island test can pass incorrectly because packages are installed #1346

Closed
opened 2011-01-31 00:13:35 +00:00 by davidsarah · 3 comments
davidsarah commented 2011-01-31 00:13:35 +00:00
Owner

The desert-island test should be testing that the contents of the deps tarball are sufficient to build without downloading. It's actually testing that those contents plus any locally installed packages, are sufficient to build without downloading.

For example, a SUMO build of 1.8.2b1, with no dependencies installed, will download setuptools_trial from PyPI in order to build zfec. That isn't caught by the desert-island builder because it has zfec locally installed.

In fact that builder has pycryptopp, mock, pyasn1, pycrypto, Nevow, foolscap, Twisted, zope.interface, simplejson, zfec, pyOpenSSL, pyutil, argparse, and zbase32 installed, so it is not using any of those from the tarball.

I believe the test should warn if it has any lines that start with "Using ".

The desert-island test should be testing that the contents of the deps tarball are sufficient to build without downloading. It's actually testing that those contents plus any locally installed packages, are sufficient to build without downloading. For example, a SUMO build of 1.8.2b1, with no dependencies installed, will download setuptools_trial from PyPI in order to build zfec. That [isn't caught by the desert-island builder](http://tahoe-lafs.org/buildbot/builders/clean/builds/2707/steps/test-desert-island/logs/stdio) because it has zfec locally installed. In fact that builder has pycryptopp, mock, pyasn1, pycrypto, Nevow, foolscap, Twisted, zope.interface, simplejson, zfec, pyOpenSSL, pyutil, argparse, and zbase32 installed, so it is not using any of those from the tarball. I believe the test should warn if it has any lines that start with "Using ".
tahoe-lafs added the
dev-infrastructure
critical
defect
1.8.1
labels 2011-01-31 00:13:35 +00:00
tahoe-lafs added this to the soon (release n/a) milestone 2011-01-31 00:13:35 +00:00

It would be really great to have a layer of isolation between the code-under-test, which in this case is the source tree and its build system, from the test-code which in this case is the buildbot itself and the other code on the host's system. The Python "virtualenv" tool might be pretty well-suited for this. I think that's what its raison d'etre is.

It would be really great to have a layer of isolation between the code-under-test, which in this case is the source tree and its build system, from the test-code which in this case is the buildbot itself and the other code on the host's system. The Python "virtualenv" tool might be pretty well-suited for this. I think that's what its raison d'etre is.
davidsarah commented 2011-08-16 04:30:14 +00:00
Author
Owner

Replying to zooko:

It would be really great to have a layer of isolation between the code-under-test, which in this case is the source tree and its build system, from the test-code which in this case is the buildbot itself and the other code on the host's system. The Python "virtualenv" tool might be pretty well-suited for this. I think that's what its raison d'etre is.

This is #1464.

Replying to [zooko](/tahoe-lafs/trac-2024-07-25/issues/1346#issuecomment-82489): > It would be really great to have a layer of isolation between the code-under-test, which in this case is the source tree and its build system, from the test-code which in this case is the buildbot itself and the other code on the host's system. The Python "virtualenv" tool might be pretty well-suited for this. I think that's what its raison d'etre is. This is #1464.
zooko added
major
and removed
critical
labels 2012-03-29 16:10:43 +00:00

We no longer provide SUMO tarballs, and the new desert-island procedure uses the pip wheel cache. Also, everything is now tested in a virtualenv. So I'm going to close this one.

We might want to have a new ticket to exercise the recommended desert-island procedure, but I'm not currently convinced we really need it.

We no longer provide SUMO tarballs, and the new desert-island procedure uses the pip wheel cache. Also, everything is now tested in a virtualenv. So I'm going to close this one. We might want to have a new ticket to exercise the recommended desert-island procedure, but I'm not currently convinced we really need it.
warner added the
fixed
label 2016-03-27 18:34:22 +00:00
warner modified the milestone from soon (release n/a) to 1.11.0 2016-03-27 18:34:22 +00:00
Sign in to join this conversation.
No Milestone
No Assignees
3 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Reference: tahoe-lafs/trac-2024-07-25#1346
No description provided.