[Twisted-Python] release momentum
Our release schedule, or lack thereof, is getting to me. I'm reluctant to deploy new releases of CVSToys because I'd be asking people to depend on an "alpha" version of Twisted. At least one person has come in to #twisted asking if there was going to be a release soon, or if he had to ask his users to get Twisted from CVS. Our last *alpha* release was was five weeks before PyCon started; there have been over five hundred commits since that time. I know we've all been busy, I know our release manager has had to devote his attention to people who will help him pay his bills -- I'm not concerned with any of that now. All I want to know is what we have to do to get another alpha release out. Tests we're having problems with: * testSignal in test_process.PosixProcessTestCase, fails regularly under buildbot for python2.2, but not 2.1. I cannot reproduce this failure. * testStor in test_ftp, intermittent failures. * ThrottlingTestCase in test_policies has intermittent failures, they may be spurious timing issues. * test_imap is under heavy development, failures changed fresh daily. * test_jelly.CircularReferenceTestCase, Jasper's bug. How many of these do we have to turn green before we can build the next alpha release? Is the jelly bug going to wait on newjelly? Since I'm the one raising a stink, I'll volunteer to manage the next release should Chris remain unavailable. (Now why did saying that start the music of foreshadowing doom in my head?) - Kevin -- The moon is waxing crescent, 6.1% illuminated, 2.3 days old.
On Fri, Apr 04, 2003 at 10:28:59AM -0800, Kevin Turner wrote:
Thanks for bringing this up. I've been wondering about it a bit myself as well.
I wouldn't be unhappy if imap4 were left out of the next release, if that release is done before 1) the protocol support is somewhat stable 2) there is a mailbox implementation in Twisted proper (I currently have one, but not in Twisted). Of course, I also wouldn't be unhappy if it were included, assuming it is in reasonable condition when someone actually does go ahead with the release.
Speaking of newjelly... Is it definite yet whether this will be backwards compatible with the current jelly yet? I assumed it would not be, but I've heard some rumors... Since our numbering scheme is currently "* glyph rolls dice.", maybe this doesn't matter, but I think it *might* be nice if the next release was 1.1.0, if jelly compatibility isn't going to be maintained. Jp -- http://catandgirl.com/view.cgi?44 -- up 15 days, 16:00, 9 users, load average: 0.00, 0.02, 0.07
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Friday, April 4, 2003, at 03:39 PM, Jp Calderone wrote:
I think we should stick with 1.0.4, since I think we should do the release ASAP, and not wait until we have enough time to test newjelly with all the applications that might need it. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (Darwin) iD8DBQE+jhedvVGR4uSOE2wRAtIxAJ0QAu+gZ38kXNiKiu922OUg0XcYTwCfdjgx oXtG5pQAwY4QnTgHS2CGGcU= =VcV6 -----END PGP SIGNATURE-----
On Fri, 4 Apr 2003 18:39:06 -0500 Glyph Lefkowitz <glyph@twistedmatrix.com> wrote:
And everyone can help by making sure they add changelog entries for any new features/APIs they added since 1.0.3. -- Itamar Shtull-Trauring http://itamarst.org/ http://www.zoteca.com -- Python & Twisted consulting
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Friday, April 4, 2003, at 01:28 PM, Kevin Turner wrote:
You definitely have a point there.
Our last *alpha* release was was five weeks before PyCon started; there have been over five hundred commits since that time.
Zow. I hadn't counted. This is pretty bad.
Neither can I. does anyone have a theory as to what's causing it?
* testStor in test_ftp, intermittent failures.
This should be fixed eventually, but it's been having problems forever. We shouldn't hold up the release based on it.
* ThrottlingTestCase in test_policies has intermittent failures, they may be spurious timing issues.
The tests are written incorrectly. They assert things that should usually be true, but aren't necessarily.
* test_imap is under heavy development, failures changed fresh daily.
JP should get these to pass, but similarly they should not hold up the release.
* test_jelly.CircularReferenceTestCase, Jasper's bug.
I'll pick this one up, since it's really my fault. I should be able to look at it on Sunday.
How many of these do we have to turn green before we can build the next alpha release? Is the jelly bug going to wait on newjelly?
The jelly bug shouldn't wait on newjelly because this involves breaking compatibility not only with old Twisted, but also with every other language implementation of PB that currently exists. These should eventually be fixed, but I think at this point that we just can't ignore compatibility in the meanwhile.
Since I'm the one raising a stink, I'll volunteer to manage the next release should Chris remain unavailable.
Let me give you some advice, from the last time I volunteered to do this. Corner Chris on IRC, get him to explain everything that needs to be done, and then take it over. Don't wait for him to agree that he won't be available :-). It's actually not that much work, now that most of it's automated. The only thing that's really time-consuming is having to re-do it all every time somebody checks in a fix, and twice when moving to final-RC to final-release. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (Darwin) iD8DBQE+jhMuvVGR4uSOE2wRAr/dAKCEw5Og+P2Ho6FnUw0mvIdfhV9LFgCgkcXw 5dOivSojnY9quQY1KQVlAQk= =jKiN -----END PGP SIGNATURE-----
On Fri, Apr 04, 2003 at 06:20:11PM -0500, Glyph Lefkowitz wrote:
Glyph, you know me too well. I'll talk to Acap on IRC >:) -- Twisted | Christopher Armstrong: International Man of Twistery Radix | Release Manager, Twisted Project ---------+ http://twistedmatrix.com/users/radix.twistd/
On Fri, Apr 04, 2003 at 10:28:59AM -0800, Kevin Turner wrote: [...]
This is a new test, but an old problem. The bug here is in the FTP server (even though we wrote the test to test the new client support for STOR, which seems to work just fine), and it hasn't changed for months. On the other hand, the failures are intermittent, and I suspect it's only a problem in the test case (it's some sort of timing issue with having the same process on either end of the socket, I think). In short: the testStor test failure shouldn't hold up the release :) -Andrew.
On Fri, Apr 04, 2003 at 10:28:59AM -0800, Kevin Turner wrote:
Thanks for bringing this up. I've been wondering about it a bit myself as well.
I wouldn't be unhappy if imap4 were left out of the next release, if that release is done before 1) the protocol support is somewhat stable 2) there is a mailbox implementation in Twisted proper (I currently have one, but not in Twisted). Of course, I also wouldn't be unhappy if it were included, assuming it is in reasonable condition when someone actually does go ahead with the release.
Speaking of newjelly... Is it definite yet whether this will be backwards compatible with the current jelly yet? I assumed it would not be, but I've heard some rumors... Since our numbering scheme is currently "* glyph rolls dice.", maybe this doesn't matter, but I think it *might* be nice if the next release was 1.1.0, if jelly compatibility isn't going to be maintained. Jp -- http://catandgirl.com/view.cgi?44 -- up 15 days, 16:00, 9 users, load average: 0.00, 0.02, 0.07
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Friday, April 4, 2003, at 03:39 PM, Jp Calderone wrote:
I think we should stick with 1.0.4, since I think we should do the release ASAP, and not wait until we have enough time to test newjelly with all the applications that might need it. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (Darwin) iD8DBQE+jhedvVGR4uSOE2wRAtIxAJ0QAu+gZ38kXNiKiu922OUg0XcYTwCfdjgx oXtG5pQAwY4QnTgHS2CGGcU= =VcV6 -----END PGP SIGNATURE-----
On Fri, 4 Apr 2003 18:39:06 -0500 Glyph Lefkowitz <glyph@twistedmatrix.com> wrote:
And everyone can help by making sure they add changelog entries for any new features/APIs they added since 1.0.3. -- Itamar Shtull-Trauring http://itamarst.org/ http://www.zoteca.com -- Python & Twisted consulting
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Friday, April 4, 2003, at 01:28 PM, Kevin Turner wrote:
You definitely have a point there.
Our last *alpha* release was was five weeks before PyCon started; there have been over five hundred commits since that time.
Zow. I hadn't counted. This is pretty bad.
Neither can I. does anyone have a theory as to what's causing it?
* testStor in test_ftp, intermittent failures.
This should be fixed eventually, but it's been having problems forever. We shouldn't hold up the release based on it.
* ThrottlingTestCase in test_policies has intermittent failures, they may be spurious timing issues.
The tests are written incorrectly. They assert things that should usually be true, but aren't necessarily.
* test_imap is under heavy development, failures changed fresh daily.
JP should get these to pass, but similarly they should not hold up the release.
* test_jelly.CircularReferenceTestCase, Jasper's bug.
I'll pick this one up, since it's really my fault. I should be able to look at it on Sunday.
How many of these do we have to turn green before we can build the next alpha release? Is the jelly bug going to wait on newjelly?
The jelly bug shouldn't wait on newjelly because this involves breaking compatibility not only with old Twisted, but also with every other language implementation of PB that currently exists. These should eventually be fixed, but I think at this point that we just can't ignore compatibility in the meanwhile.
Since I'm the one raising a stink, I'll volunteer to manage the next release should Chris remain unavailable.
Let me give you some advice, from the last time I volunteered to do this. Corner Chris on IRC, get him to explain everything that needs to be done, and then take it over. Don't wait for him to agree that he won't be available :-). It's actually not that much work, now that most of it's automated. The only thing that's really time-consuming is having to re-do it all every time somebody checks in a fix, and twice when moving to final-RC to final-release. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (Darwin) iD8DBQE+jhMuvVGR4uSOE2wRAr/dAKCEw5Og+P2Ho6FnUw0mvIdfhV9LFgCgkcXw 5dOivSojnY9quQY1KQVlAQk= =jKiN -----END PGP SIGNATURE-----
On Fri, Apr 04, 2003 at 06:20:11PM -0500, Glyph Lefkowitz wrote:
Glyph, you know me too well. I'll talk to Acap on IRC >:) -- Twisted | Christopher Armstrong: International Man of Twistery Radix | Release Manager, Twisted Project ---------+ http://twistedmatrix.com/users/radix.twistd/
On Fri, Apr 04, 2003 at 10:28:59AM -0800, Kevin Turner wrote: [...]
This is a new test, but an old problem. The bug here is in the FTP server (even though we wrote the test to test the new client support for STOR, which seems to work just fine), and it hasn't changed for months. On the other hand, the failures are intermittent, and I suspect it's only a problem in the test case (it's some sort of timing issue with having the same process on either end of the socket, I think). In short: the testStor test failure shouldn't hold up the release :) -Andrew.
participants (6)
-
Andrew Bennetts -
Christopher Armstrong -
Glyph Lefkowitz -
Itamar Shtull-Trauring -
Jp Calderone -
Kevin Turner