Rework purebrowser packaging
Open, NormalPublic

Description

Tue 18 21:50 < zlatan> just take latest ff package from debian, rebrand it (name/artwork), add those two extensions (xul packages for ublock origin and https everywhere), disable the add-ons page with same thing that current purebrowser is using, 
default to ddg search engine, see to disable google/yahoo/and other invasive search engines, have correct 
replaces/conflicts etc
  1. Import purebrowser from firefox-esr:45.4.0esr-2
  2. Add Hema's work as a branch or somehow diff the two
  3. Cherry-pick the relevant/kept changes
  4. Fix actual issues! \o/

Reimported all available dscs. Not all are available for some reason.

While it was DDG before, in the meantime somehow that became a *not* true statement.
https://tracker.pureos.net/T156

Purebrowser is currently out of date as rebuilding the latest esr results in extremely strange rendering artifacts such as completely missing icons and lines, etc. until I mouseover:

https://i.imgur.com/LRaquLX.jpg

A regular firefox-esr does not suffer from this problem, so I am (temporarily!) stumped as to why this happens.

jeff added a subscriber: jeff.EditedOct 3 2017, 11:49 PM

Pardon my silly question, but as the latest release of Firefox ESR is version 52.4, could we take the opportunity to leap to that instead of being stuck at the 1.5-years-old 45 series, and lock ourselves into version 52 for the foreseeable future? (I would tend to think we should not jump to regular Firefox 57+ for a while, because the forceful migration to webextensions might cause us untold misery)

habs added a subscriber: habs.Nov 18 2017, 1:04 AM

Purebrowser is currently out of date as rebuilding the latest esr results in extremely strange rendering artifacts such as completely missing icons and lines, etc. until I mouseover:

https://i.imgur.com/LRaquLX.jpg

A regular firefox-esr does not suffer from this problem, so I am (temporarily!) stumped as to why this happens.

I just built firefox-esr_52.4.0esr-2_amd64.deb from the Debian testing sources on an up-to-date PureOS install and it works fine for me, not experiencing that issue. The way the repo I can see is set up (a fork without patches), it seems like it would be extremely difficult to port *all* of the changes line-for-line to the new version -- but I should be able to just port over the branding (into browser/branding/official) and the extensions.

Trisquel brands their Firefox clone with these two scripts and this folder of data:

https://devel.trisquel.info/trisquel/package-helpers/blob/550bed8e8a39a20cb81add1df97a34af4c3b2c7b/helpers/config
https://devel.trisquel.info/trisquel/package-helpers/blob/550bed8e8a39a20cb81add1df97a34af4c3b2c7b/helpers/make-firefox
https://devel.trisquel.info/trisquel/package-helpers/tree/550bed8e8a39a20cb81add1df97a34af4c3b2c7b/helpers/DATA/firefox

I just built firefox-esr_52.4.0esr-2_amd64.deb from the Debian testing sources on an up-to-date PureOS install and it works fine for me, not experiencing that issue.

Right, to be clear, it's only once you change the branding (specifically *just* the Debian source package name!) you start to see this issue.

It seems *only* renaming package will fail (EDIT: or so I guess - it regenerates control file dropping build-dependency on gcc-6) due to our change of target suite, and (at least, haven't tried yet) this patch is needed:

}diff --git a/debian/upstream.mk b/debian/upstream.mk
index 0cb72b1..31bdaab 100644
--- a/debian/upstream.mk
+++ b/debian/upstream.mk
@@ -42,7 +42,7 @@ DIST = unknown
 ifneq (,$(filter experimental,$(DEB_DISTRIBUTION)))
 DIST = experimental
 endif
-ifneq (,$(filter testing% buster% unstable sid,$(DEB_DISTRIBUTION)))
+ifneq (,$(filter testing% buster% unstable sid green landing,$(DEB_DISTRIBUTION)))
 DIST = buster
 endif
 ifneq (,$(filter bpo% deb%,$(DEBIAN_RELEASE_EXTRA)))

As in FTBFS? Wfm, but also do note that this package uses a control.in...

No as in not-how-Debian-maintainer-intended-for-sid - e.g. causing rendering artifacts.

Yes, I am aware of the control.in - that's where I applied other tweaks.

Here is my starting point for a PureBrowser packaging - tested briefly (loading https://www.debian.org/ and https://apple.com/ without visual artifacts): https://code.puri.sm/pureos/firefox-esr/commits/master

Previus comment updated to clarify the sloppy and confusing use of "fail".

@chris.lamb Please do tell me if you would like me to finalize the packaging that I have succesfully test-built.

Alternatively, please tell me how - if at all - you imagine that I can maybe help here.
I feel that - despite your several clues both here and in emails - I fail to understand in what way I can help this further, without the risk of doing duplicate work of yours.

@jonas.smedegaard First, perhaps I am accidentally reading something in your messages that is not there but I truly hope there is no animosity or something other that purely technical discussion here? If there is, then let's fix that and move on. :)

In terms of the next steps, can you link to your built packages for potential testing and the repo you used to create them?

Some quick questions would be: a) in which environment did you build them (if you aren't shipping a .buildinfo, that is!) and b) how much testing of the upgrade you have done (no worries if "none", just curious atm.)

Thanks!

jonas.smedegaard added a comment.EditedDec 13 2017, 9:11 AM

Reason for my cautious wording of comment T147#4606 was concern for you: You have fought this issue for a long time and mentioned frustration about it on irc recently. I want to help but am genuinely uncertain what you consider helpful here.

I can take over this task if you want.

Regarding built packages, I can upload those but will then need to figure out where in Purism infrastructure we most suitably host them (190 binary packages, 440 MB).

Regarding source repo, it is here: https://code.puri.sm/pureos/firefox-esr/commits/master.
(or do you mean something else? We already discussed that git by email December 6th and the git repo was referenced in my above comment T147#4595).

Regarding build environment, I use cowbuilder.

Regarding testing, I briefly loaded a few websites including https://www.debian.org/ and https://apple.com/ and did not notice any visual artifacts (also described in above comment T147#4595).

concern for you: You have fought this issue for a long time and mentioned frustration about it on irc recently.

Thank you. I was feeling really quite burnt out on this issue, both in terms of technically in hitting a wall, but mostly mentally. After making real progress on PureOS endorsement recently (which also took a while), it highlighted to me just how burnt out I was on this... Yes, please take over this issue. Happy to assist, naturally, but I would immediately feel so so much more energised.. :/ Thank you.

Ok, I take over from here. Pick something fun to work on (or take a break, I am sure noone minds!).

I will post further updates to this bugreport - for anyone to follow, not you in particular: Feel free to grab some popcorn and follow my struggle, and feel free to chime in with any and all suggestions you might have, but don't feel obliged to do so. I am super happy that what I take over is a clean re-base from Debian source, rather than the challenging stuff you dealt with - I am still amazed you got that as much into order as you did :-)

nicole added a subscriber: nicole.Jul 29 2018, 8:16 PM
jonas.smedegaard removed jonas.smedegaard as the assignee of this task.Aug 20 2018, 7:27 AM
jonas.smedegaard added a subscriber: jonas.smedegaard.
jonas.smedegaard removed a subscriber: jonas.smedegaard.

Add Comment