release 19.8.0

[benoitc]

As I would like to release a 19.8 version this week. I don't see anything preventing it right now. Tests pass here but let me know if I missed something.

[benoitc]

cc @tilgovi @berkerpeksag

[berkerpeksag]

I would like to get the following PRs merged for 19.8:

• Fallback to the original behavior if prefix is not available (#1068) #1499 (I can address my own comments if we decide to merge this one) (closed as wontfix)
• Deprecate gaiohttp worker and document alternatives #1569 (will merge this tonight)
• Fix umask option bug(issues#1622) #1632 (reviewed)

[benoitc]

@berkerpeksag i commented/approved then. Let me know if you need some help as I may have some cycles tomorrow :)

[tilgovi]

I have open PRs. Do we want to merge any of them?

• Set CWD and Python path before and after config #1481 (allow Python config files to set chdir)
• Support Dictionary Config setting. #1602 (I can address comments from @berkerpeksag)

[berkerpeksag]

Both PRs good to have if we have time to get them in.

[tilgovi]

Mine are in. Can I help with reviewing anything?

[berkerpeksag]

#1499 is the only remaining one. I've addressed my review comments.

[benoitc]

@berkerpeksag i've commented also, what do you think about it?

[benoitc]

@berkerpeksag i didn't get an answer about #1499 . I'm dubious about putting back a eature that have been removed a long itme ago. We should see the pro/cons about it. Or at least makng sure it has no impact.

I think however that before releasing i will disable by defaut the SO_REUSEPORT option of the socket. Thoughts?

[RonRothman]

@benoitc I'm not sure I understand - why would we want to change the default for SO_REUSEPORT based on that PR?

(We do rely on SO_RESUSEPORT to help us efficiently serve tens of thousands of requests/sec, so I'm careful around changes to it.)

[tilgovi]

Having SO_REUSEPORT as a default seems to confused some people, especially those new to Gunicorn or using it in a development environment. It makes it possible to start a new Gunicorn without killing the old one, resulting in multiple versions of an application running. This is not a very friendly experience for a first time user who wonders why their application serves the right response only half of the time.

We will definitely not remove the functionality even if we change the default.

[berkerpeksag]

@berkerpeksag i didn't get an answer about #1499 . I'm dubious about putting back a eature that have been removed a long itme ago. We should see the pro/cons about it. Or at least makng sure it has no impact.

Looking at #1499 again, I'm fine with closing it as 'wontfix'. Making it compatible with Gunicorn 19.4+ is pretty trivial:

-            '-c', 'airflow.www.gunicorn_config'
+            '-c', 'python:airflow.www.gunicorn_config'

[benoitc]

on my todo list before any release I have the following:

• Make SO_REUSEPORT Optional #1669 (do not enable SO_REUSEPORT usage in default)
• Support wsgi.input_terminated flag #1653

i will try to have something for friday now

[robyoung-digital]

Is there any more news on this?

[berkerpeksag]

I've just merged PR #1669 and closed PR #1499 so I think the only blocker is now #1653.

[berkerpeksag]

I think we also need to document the SO_REUSEPORT behavior change (PR #1669) in release notes.

[robyoung-digital]

Wow, thanks for the quick response. 💯

[Allu2]

Knowing asking ETA's is the greatest sin of all, are there any news on progress of #1653 which seems to be left blocking the release?

[berkerpeksag]

@benoitc @tilgovi should we leave #1653 to 19.9 and release 19.8 as is? There are some important fixes in master and there are some questions from users on ETA for 19.8 in several issues (for example, #1058) so I'm wondering if we should skip #1653 for now.

[berkerpeksag]

There's an almost ready PR for #1324:

• PR Issue #1324: Fix passing pyc files to --config in Python 3.3+ #1696

[tilgovi]

I am fine to skip #1653

[benoitc]

I don't see any reason to not release more often :) If I don't came with a patch by today let's release tomorrow in the morning.

About that pyc stuff, i would be for fixing it in the current release and not adding any new option. I would rather deprecate this feature as it breaks the contract of auto reloading the config and such things.

[Allu2]

Pypi seems to be still at 19.7.1 https://pypi.python.org/pypi/gunicorn
As a side thought could the README.md be made to have stable version number?
going trough pypi to check what is the current release feels bit funky, or perhaps I'm missing some obvious version number displayed somewhere 😕

[JonathanHuot]

Sorry to bump, any news about the 19.8 release ? I'm eager to play with dictConfig :-)

[hugovk]

(@Allu2 Please see PR #1727 to show the current release's version number in the README)

[unexceptable]

Any update on this? I'm looking forward to it since it fixes some issues with --reload in alpine containers and would make our dev environments much nicer!

[tilgovi]

I'll prepare a release. @berkerpeksag @benoitc good for you?

[berkerpeksag]

@tilgovi +1 from me.

[Allu2]

Out of curiosity, what kind of process making a gunicorn release is?
This release seems to have been postponed for the past 5 months with more then one "next (day | week | soon)"

[tilgovi]

@Allu2

• Update changelog
• Tag and upload
• Make release on GitHub

It's not hard, but all of us do this on a volunteer basis.

[tilgovi]

I've updated the changelog and I believe everything is prepped for release.

I'm going to sleep and will cut the tag in the morning, when I can be around in case of any mistake that requires a followup patch release, but I don't expect any issues.

[tilgovi]

19.8.0 is released!

Please comment here with any feedback or issues!

[tilgovi]

We're going to aim for more frequent releases going forward. Your pull requests and reviews help. If you're interested in being a maintainer, let me know!

[benoitc]

@tilgovi thank you! and agree :) sorry to not have answered soon I was traveling all the month