1096244 - Mozregression installs too many unnecessary dependencies

Status: RESOLVED FIXED

(Testing :: mozregression, defect)

Description

[Ed Morley [:emorley]]

In a clean virtualenv:

$ pip install mozregression
<snip>
$ pip list
amqp (1.4.6)
anyjson (0.3.3)
BeautifulSoup (3.2.1)
blessings (1.6)
futures (2.2.0)
httplib2 (0.9)
kombu (3.0.23)
manifestparser (0.8)
mozcommitbuilder (0.4.11)
mozcrash (0.14)
mozdevice (0.43)
mozfile (1.1)
MozillaPulse (1.0)
mozinfo (0.7)
mozInstall (1.10)
mozlog (2.7)
moznetwork (0.24)
mozprocess (0.21)
mozprofile (0.22)
mozregression (0.25)
mozrunner (6.5)
mozversion (1.0)
pip (1.5.6)
pulsebuildmonitor (0.90)
python-dateutil (1.5)
pytz (2014.9)
requests (2.4.3)
setuptools (3.6)
simplejson (3.6.5)


mozregression doesn't need all of those, we're just pulling them in via mozprofile, mozrunner etc

Comment 1

[William Lachance (:wlach)]

(In reply to Ed Morley [:edmorley] from comment #0)
> In a clean virtualenv:
> 
> $ pip install mozregression
> <snip>
> $ pip list
> amqp (1.4.6)
> anyjson (0.3.3)
> BeautifulSoup (3.2.1)
> blessings (1.6)
> futures (2.2.0)
> httplib2 (0.9)
> kombu (3.0.23)
> manifestparser (0.8)
> mozcommitbuilder (0.4.11)
> mozcrash (0.14)
> mozdevice (0.43)
> mozfile (1.1)
> MozillaPulse (1.0)
> mozinfo (0.7)
> mozInstall (1.10)
> mozlog (2.7)
> moznetwork (0.24)
> mozprocess (0.21)
> mozprofile (0.22)
> mozregression (0.25)
> mozrunner (6.5)
> mozversion (1.0)
> pip (1.5.6)
> pulsebuildmonitor (0.90)
> python-dateutil (1.5)
> pytz (2014.9)
> requests (2.4.3)
> setuptools (3.6)
> simplejson (3.6.5)
> 
> 
> mozregression doesn't need all of those, we're just pulling them in via
> mozprofile, mozrunner etc

Most of the really silly dependencies are being sucked in via mozcommitbuilder, which I'm not sure even works anymore (probably not). I wonder if we should just nuke that.

Comment 2

[Julien Pagès (:parkouss)]

Maybe we can replace the actual call to mozcommitbuilder with some help to indicate how to use the mozcommitbuilder command line (with our good/bad revs), EG:

[pip install mozcommitbuilder]
mozcommitbuilder -g <good-rev> -b <bad-rev> [-j 8]

This way we remove the need of mozcommitbuilder, but still we help user if they want to use it.

Comment 3

[Julien Pagès (:parkouss)]

Hey,

I would like to remove the mozcommitbuilder dependency in mozregression. As a workaround, I propose what is indicated in comment 2.

But I am not really a final user of mozregression. What do you think, is that ok for you if mozregression does not integrate mozcommitbuilder but instead indicate how to use it ? Do you use it anyway ?

@William
Also, what I really like to do is remove all the bugs that are mozcommitbuilder related from the mozregression bugs. How could we achieve that ? create a new bugzilla component ? Or looking at your comment 1, maybe mozcommitbuilder is just dead, and we have to find another solution than using mozcommitbuilder (ie just remove all these obsolete bugs) ?

Comment 4

[Ed Morley [:emorley]]

I'll defer to Will on this one :-)

Comment 5

[William Lachance (:wlach)]

I suspect we may just want to nuke mozcommitbuilder, unless you or someone else wants to maintain it. I really don't have the bandwidth to help maintain this tool which isn't very widely used at all. My experience has been that the inbound regression ranges are usually tight enough that there's not a lot of value in this tool anymore (mozcommitbuilder was created before mozregression had inbound support).

Comment 6

[Julien Pagès (:parkouss)]

Ok, so let's remove the mozcommitbuilder dependency and we won't provide support to use it from mozregression. If someone complains, we can still think about it at this time - maybe by integrating some of the code of mozcommitbuilder inside mozregression, or something else.

This is now quite a simple bug to resolve - I will put this as a good first bug and set myself as a mentor, and once finished I will resolve all mozcommitbuilder related bugs as wontfix, if no ones has objections here. :)

Comment 7

[Julien Pagès (:parkouss)]

Attachment: remove mozcommitbuilder dependency

I changed my mind and will resolve the bug myself - it is really simple.

I know that Elbart did not answered yet about this, maybe we should wait for him? I let you decide William, whether or not you are sure to remove mozcommitbuilder from mozregression. :)

Comment 8

[William Lachance (:wlach)]

Comment on attachment 8544569 [details] [review]
remove mozcommitbuilder dependency

Thanks for doing this. I think this is the right thing to do. If there's serious demand / complaints we can consider reviving mozcommitbuilder (or building a replacement). 

I'm going to clear the needinfo for elbart, as I think we have enough data to go ahead.

Comment 9

[Julien Pagès (:parkouss)]

Ok merged. I'm going to close this bug as RESOLVED, and all the mozcommitbuilder related other bugs as WONTFIX. This will give us some space on bugzilla. :)