Proposed changes to supported build configurations (tiers)
Benjamin Smedberg
has some build configurations which don't match our actual practices. I'd
like to make the following changes:
Add to tier 1:
* Linux/x86-64
* OSX/x86-64
Change the description of tier 2:
"Tier 2 platforms are platforms that the Mozilla community believes are
important to maintain, but breakage in these platforms do not require
immediate backout. Developers who break these platforms should work with
port maintainers to fix any problems, and may need to back out if a solution
cannot be found."
Add to tier 2:
* Android Linux/ARM
Move to tier 3:
* OpenSolaris
* Windows CE/Windows Mobile
* OS/2
* win32/mingw gcc
If there are objections to these changes, please let me know ASAP.
--BDS
Ted Mielczarek
<benj...@smedbergs.us> wrote:
> Add to tier 2:
>
> * Android Linux/ARM
I'd suggest we also add Windows/x86-64 (msvc) to Tier 2, since AFAIK
we're not shipping it for Firefox 4, but we do have builds/tests on
the Firefox tinderbox.
-Ted
Mike Beltzner
> The page at https://developer.mozilla.org/en/Supported_build_configurations has some build configurations which don't match our actual practices. I'd like to make the following changes:
>
> Add to tier 1:
>
> * Linux/x86-64
> * OSX/x86-64
Do we want to add Maemo Linux/ARM here? As far as I understand, that's the level of support that we've been giving it lately.
> Change the description of tier 2:
>
> "Tier 2 platforms are platforms that the Mozilla community believes are important to maintain, but breakage in these platforms do not require immediate backout. Developers who break these platforms should work with port maintainers to fix any problems, and may need to back out if a solution cannot be found."
>
> Add to tier 2:
>
> * Android Linux/ARM
I suspect we'll be moving OS X/ppc to this tier, so I might as well tack that on to this list of proposed changes.
cheers,
mike
Benjamin Smedberg
> On 2010-08-25, at 9:33 AM, Benjamin Smedberg wrote:
>
>> The page at
>> https://developer.mozilla.org/en/Supported_build_configurations has
>> some build configurations which don't match our actual practices. I'd
>> like to make the following changes:
>>
>> Add to tier 1:
>>
>> * Linux/x86-64 * OSX/x86-64
>
> Do we want to add Maemo Linux/ARM here? As far as I understand, that's
> the level of support that we've been giving it lately.
I don't know. We've discussed it several times in the past, and the
consensus was that we couldn't do that until we had tryserver tests. I
believe that we have tryserver builds now, but not tryserver unit or
performance tests. We could certainly make *building* on Maemo a tier-1
issue. Similarly with Android.
> I suspect we'll be moving OS X/ppc to this tier, so I might as well tack
> that on to this list of proposed changes.
I'll do that whenever it disappears from the Firefox tinderbox, sure.
--BDS
Philip Chee
> Move to tier 3:
>
> * OpenSolaris
Isn't OpenSolaris D.O.A.? IIRC the entire OpenSolaris board resigned en
mass.
Phil
--
Philip Chee <phi...@aleytys.pc.my>, <phili...@gmail.com>
http://flashblock.mozdev.org/ http://xsidebar.mozdev.org
Guard us from the she-wolf and the wolf, and guard us from the thief,
oh Night, and so be good for us to pass.
Patrick Finch
On 25/08/2010 17:03, Philip Chee wrote:
> On Wed, 25 Aug 2010 09:33:11 -0400, Benjamin Smedberg wrote:
>
>> Move to tier 3:
>>
>> * OpenSolaris
>
> Isn't OpenSolaris D.O.A.? IIRC the entire OpenSolaris board resigned en
> mass.
>
> Phil
The OpenSolaris project indeed looks moribund, but I think Oracle (Sun)
will continue to develop Solaris.
Patrick
--
Patrick Finch
Mozilla
pat...@mozilla.com
Mobile: +46 768 444 833
Office: +1 650 903 0800 ext. 340
Twitter: @patrickf
IM: patric...@gmail.com
Ginn Chen
And the breakage is not show up in tinderbox Firefox page because it is only built for ppc but not tested.
I think we can say it is already "disappears from the Firefox tinderbox".
I'm very sad to see OpenSolaris is moved to tier 3.
It might be true that it no longer fits the new description of tier 2, "Mozilla community believes are important".
But my feeling is hurt because "actively maintained by port maintainers" is no longer recognized.
Regards,
Ginn
Stuart Parmenter
> The page athttps://developer.mozilla.org/en/Supported_build_configurations
I believe that we need to move both Android and Maemo to Tier 1. They
are both primary focus platforms for the Mozilla community and need to
be treated as such.
stuart
Mike Beltzner
> I believe that we need to move both Android and Maemo to Tier 1. They
> are both primary focus platforms for the Mozilla community and need to
> be treated as such.
I'd proposed moving Maemo to Teir 1 and Android to Tier 2, and think that makes sense until we've shipped a beta of the Android version of the browser, but agree that we should continue to aggressively work through failures on those Tier 2 platforms when they crop up.
Moving to Tier 1 means that users cannot check in unless bustage is fixed and test failures are starred. Right now the Maemo trees have a bunch of unstarred oranges; is that usual?
cheers,
mike
Shawn Wilsher
> Moving to Tier 1 means that users cannot check in unless bustage is fixed and test failures are starred. Right now the Maemo trees have a bunch of unstarred oranges; is that usual?
which I think was pointed out as to why it cannot be tier 1 yet.
Cheers,
Shawn
Peter Weilbacher
> The page at
> https://developer.mozilla.org/en/Supported_build_configurations has some
> build configurations which don't match our actual practices.
[...]
> Move to tier 3:
> * OS/2
[...]
> If there are objections to these changes, please let me know ASAP.
Well, OS/2 is maintained, even though I am no longer taking part, 3-4 people
do work on it pretty actively. Although help we have gotten from mozilla.com
employees to fix OS/2 breakages outside JS was often not worth mentioning, I
don't think OS/2 falls in the same category as BeOS, QNX, or HPUX.
Peter,
Robert Sayre
>
>
> I believe that we need to move both Android and Maemo to Tier 1. They
> are both primary focus platforms for the Mozilla community and need to
> be treated as such.
This proposal is not tractable.
The first problem is that these builds are building a different product
than the Firefox builds, with some substantive differences (e.g. e10s).
It's not reasonable to expect developers be productive when they have to
land on 7 platforms across two products. In particular, performance
regressions and improvements will have many more confounding variables.
The second problem is that the level of automated testing for the mobile
product is completely inadequate. Tier 1 platforms are tested
exhaustively, in an automated fashion. You make what you measure.
The third problem is that the platforms themselves are inadequately
specified. Mobile platforms are more heterogeneous than x86 products, so
saying "Android ARM" is not saying much. The mobile team should have a
written, forward-looking platform support plan. That would benefit the
mobile product, and the project as a whole.
The fourth problem is the style of integration that mobile builds
employ. They pull the tip of mozilla-central as well as the tip of the
mobile repository. We have similar builds displayed on the TM tree, as
the JS team is more likely than most to have architecture-specific
problems with builds. These builds often fail because of problems in the
mobile repository. Problems happen, so that's ok, but it is an indicator
that mobile integration work should be taking place in a separate
repository. That's what TM does, and it works well.
The fifth problem is that the mobile build system is much flakier than
the desktop equivalent. Every time I've looked into an issue, I've found
that it is not doing what it claims to be, even if it happens to be working.
Making Android and Maemo a Tier 1 platforms on mozilla-central would
shift mobile integration work to engineers working on desktop Firefox,
but the outcome won't be a good one, for the reasons listed above.
Since I am in favor of arrangements that will actually work, I propose
the following changes:
1.) Tier 1: Firefox running on Ubuntu or Fedora, on a Tegra2 ARM
machine, with the full suite of regression tests running.
2.) Mobile builds from a repository that merges regularly with
mozilla-central, instead of directly on mozilla-central.
This proposal provides much more extensive test coverage for ARM, and
allows both desktop and mobile firefox more flexibility than exists now.
That last part sounds a little obsequious, but I mean it. The
tracemonkey and mozilla-central trees both benefit from the fact that
most serious problems only occur in one of the two repositories, rather
than both.
We have distributed version control. We should use it. Separate
repositories are one honking great idea -- let's do more of those!
- Rob