864751 - regular profiles and DEBUG=1 profiles should live in two different directories

Status: RESOLVED FIXED

(Firefox OS Graveyard :: General, defect)

Description

[Anthony Ricaud (:rik)]

We often switch between a profile that we use in Firefox Desktop with the developer tools and a profile that we send on the device.

Right now, you need to |rm -rf profile| every time. I think it would be saner and more productive to use two different directories for this.

Comment 1

[Julien Wajsberg [:julienw]]

We can't actually do that this way, because there are existing scripts that would break if we'd do such a change.

However, Vivien and I discussed about that lately and we thought the best solution could be to add a "PROFILE=" env variable, that would control the directory name where the profile will be stored. This way, we can easily write new scripts without breaking existing scripts.

Comment 2

[Julien Wajsberg [:julienw]]

BTW we also thought of having separate profiles for tests and developing in the browser, which is why being able to chose the profile directory name is useful too.

Comment 3

[Anthony Ricaud (:rik)]

Do we have any script that uses DEBUG=1 ? I wouldn't expect that since it breaks the device.

If there's no script using it, we can keep the current default profile/ without DEBUG=1 and write to a new directory with DEBUG=1.

Comment 4

[Julien Wajsberg [:julienw]]

Every person that runs the unit tests in a scripted way have scripts that do that.

I do, for example, but I'm quite sure releng do that too for tbpl. I don't want to change that now that we're quite close to have tbpl working.

Oh, and the Travis script (that should be easy to fix too, granted).

Comment 5

[Tim Guan-tin Chien [:timdream] (please needinfo)]

Tony, Julien, we are like minded! I totally agree we should have a configurable GAIA_PROFILE env variable.

Comment 6

[Dave Hylands [:dhylands]]

Ideally, I'd like to see a separate tree for all generated files, not just the profile (so similar in concept to objdir-gecko for the gecko tree).

This allows the source tree to remain "clean".

Comment 7

[Julien Wajsberg [:julienw]]

Good idea Dave ! But I think this could be in another bug. Would you please file one ?

Comment 8

[Kevin Grandon :kgrandon]

This is a good idea and I've wanted it for a while. I'm going to implement something, and I don't think it would be too hard to fix up any broken scripts and find the ones to fix after a few emails to the mailing lists.

Comment 9

[Kevin Grandon :kgrandon]

Attachment: Github pull request pointer

Comment 10

[Kevin Grandon :kgrandon]

Comment on attachment 765199 [details]
Github pull request pointer

Adding a few people for review - anyone feel free to take a look and comment. Thanks.

Comment 11

[Julien Wajsberg [:julienw]]

Comment on attachment 765199 [details]
Github pull request pointer

finished the review on github

thanks !

Comment 12

[Anthony Ricaud (:rik)]

You'll need to change travis.sh also to use the right profile folder.

Comment 13

[Julien Wajsberg [:julienw]]

AFAIK that's already in the patch ;)

Comment 14

[Kevin Grandon :kgrandon]

Comment on attachment 765199 [details]
Github pull request pointer

Ok Julien - I think this is ready for another review. Thanks!

Comment 15

[Kevin Grandon :kgrandon]

Comment on attachment 765199 [details]
Github pull request pointer

Er, give me one minute. Might need to fix one thing.

Comment 16

[Kevin Grandon :kgrandon]

Comment on attachment 765199 [details]
Github pull request pointer

Nevermind, it was a bad alias. This should *really* be ready for another pass :)

Comment 17

[Julien Wajsberg [:julienw]]

Comment on attachment 765199 [details]
Github pull request pointer

last round of comments I guess ! :)

Comment 18

[Kevin Grandon :kgrandon]

Comment on attachment 765199 [details]
Github pull request pointer

Awesome - last requests have been implemented. Thanks for the reviews!

Comment 19

[Julien Wajsberg [:julienw]]

Comment on attachment 765199 [details]
Github pull request pointer

r=me

thanks !

Comment 20

[Julien Wajsberg [:julienw]]

a=npotb means we should push this to v1-train too, right ?

Comment 21

[Anthony Ricaud (:rik)]

This is not "not part of the build". It is the build system… But I'd like to see it on as many branches as possible.

Comment 22

[Kevin Grandon :kgrandon]

Attachment: Pull request part 2

This is part 2. I must've made a mistake during rebasing to master. This was a simple typo fix which brought the code into the R+'d expected state.

Comment 23

[Kevin Grandon :kgrandon]

This has landed, but unfortunately there are two merge commits. A second smaller fix seemed better than backing everything out and generating the same amount of commits.

Part 1: https://github.com/mozilla-b2g/gaia/commit/16178947f4f68a371226d284021f9a4461c28675
Part 2: https://github.com/mozilla-b2g/gaia/commit/1c4da9d4a0b2bae459d3d199b8ca7aa1b96725ad

Comment 24

[Julien Wajsberg [:julienw]]

So should we land this on v1-train without leo+ or approval, with a=build ?