1214192 - Add the ability to launch specific releases

Status: RESOLVED FIXED

(Testing :: mozregression, defect)

Description

[Ryan VanderMeulen [:RyanVM]]

mozregression --launch works nicely if you happen to know the date of the build you want, but not if you just want to launch a specific version. I've resorted to either downloading the various Portable Edition releases for easy access or running |mozregression --good-release 44| to error out and get the list of releases with their corresponding dates that I can then feed into --launch.

It would be nice if we had a --launch-release option where the version could be directly specified.

Comment 1

[Julien Pagès (:parkouss)]

Yeah, sounds interesting.

Note that you can run an inbound build (for firefox, defaulting on mozilla-inbound branch) by giving a changeset to the --launch argument.

Comment 2

[Julien Pagès (:parkouss)]

Mikeling, maybe you are interested to work on this bug ? The idea would be to allow --launch to accept a release.

My first idea would be to check if the argument is a known release, and in that case we translate that to the right date and go on with that date to launch the build.

Comment 3

[MikeLing]

Yep, I would happy to work on this issue! Thank you for let me know this :)

Comment 4

[MikeLing]

Attachment: PR for bug 1214192

Hey Julien, I meet some HTTPSConnection problem when I test my PR, I do meet a same problem when I work on offer a way to launch a build by data or revision. 

Actually, the most thing I have do in this PR is convert release number to data and pass it to "launch_nightly". And you require that we need to launch inbound also in this issue, which is I haven't ideal about that. So, I would like to ask feedback and told you my problem first, and hear your opinion about it :)

It's my error message BTW, https://pastebin.mozilla.org/8849674 Thank you for your help!

Comment 5

[Julien Pagès (:parkouss)]

Comment on attachment 8675412 [details] [review]
PR for bug 1214192

Well, I was thinking about just try to handle the release with the existing --launch flag, ie:

mozregression --launch 33

I would prefer not to add a new flag. Note that another possibility would be to use something like:

mozregression --launch release:33

But it seems like an overkill, I don't see any reason why a release number would conflict with a date or a changeset for now. Maybe ask :wlach to have his opinion, else I would say we just go for the first option (--launch $RELEASE).

Looking at your error, it seems that you are using a low timeout of 10s (default is 30). try to increase it using "--http-timeout 40" (or in the config file: http_timeout = 40). I hope that will help. :)

Comment 6

[Julien Pagès (:parkouss)]

Oh, inbound launch is already implemented, this bug is only for releases - https://github.com/mozilla/mozregression/pull/331/files#diff-a753870fc2c6fa4864480ab7b1c1cd57R483.

Comment 7

[Julien Pagès (:parkouss)]

Comment on attachment 8675412 [details] [review]
PR for bug 1214192

Almost there, but still not working:

> mozregression --launch ceb07c2b56f92e7674a5f54ecdb3e04d9262f2a6
>  0:00.34 LOG: MainThread WARNING You are using mozregression version 1.0.1, however version 1.0.4 is available.
>  0:00.34 LOG: MainThread WARNING You should consider upgrading via the 'pip install --upgrade mozregression' command.
> Traceback (most recent call last):
>   File "/home/jp/.virtualenvs/mozregression/bin/mozregression", line 9, in <module>
>     load_entry_point('mozregression==1.0.1', 'console_scripts', 'mozregression')()
>   File "/home/jp/dev/mozregression/mozregression/main.py", line 268, in main
>     config.validate()
>   File "/home/jp/dev/mozregression/mozregression/cli.py", line 489, in validate
>     options.launch = parse_date(date_of_release(options.launch))
>   File "/home/jp/dev/mozregression/mozregression/releases.py", line 62, in date_of_release
>     return releases()[int(release)]
> ValueError: invalid literal for int() with base 10: 'ceb07c2b56f92e7674a5f54ecdb3e04d9262f2a6'

This is basically because of:
> try:
>     ...
> except KeyError, ValueError:
>     ...

This is not catching both KeyError and ValueError - this is catching only KeyError, and the exception instance is stored in a variable named ValueError.

flake8 warned about that: https://travis-ci.org/mozilla/mozregression/builds/86805624.

The right syntax you want to use is:
> try:
>     ...
> except (KeyError, ValueError):
>     ...

I'll add a couple of comments on github too.

Please ask again for a review when you fixed that. :) Note that you PR should be green on github since you are adding unit tests, so pay attention to the check panel in https://github.com/mozilla/mozregression/pull/334.

Thanks!

Comment 8

[MikeLing]

Attachment: PR for bug 1214192

Hope this one can works well :)

Comment 9

[Julien Pagès (:parkouss)]

Comment on attachment 8677910 [details] [review]
PR for bug 1214192

Yep thanks Mikeling! Seems to work well, and github hooks are all green. :)

I'm glad to see that feature, glad to see some tests for this too.

Merged in https://github.com/mozilla/mozregression/commit/71dd468dab33850310ac8d56b28514fd76c86494.

Comment 10

[Julien Pagès (:parkouss)]

I'm going to release a mozregression 1.1.0 with this (and othr good stuff), waiting a bit for archive.mozilla.org to be back first I think.