PSA: ARM toolchain changes, avoid sync'ing to ToT

6 views
Skip to first unread message

Anush Elangovan(அனுஷ்)

unread,
May 10, 2012, 12:34:04 PM5/10/12
to Chromium OS dev
tl;dr If you are working on any ARM target it is best to avoid syncing
to ToT right now. There are some toolchain deployment issues being
worked out. Please wait for an update to this thread when we have
addressed the issues.

Anush Elangovan(அனுஷ்)

unread,
May 10, 2012, 3:35:19 PM5/10/12
to Chromium OS dev
Once the tree goes green you should be able to repo sync and
everything should just work.

Your first build may move your /build/arm-board/ out and rebuild it
cleanly (the upgrade script will notify you about it).

Don Garrett

unread,
May 10, 2012, 4:45:24 PM5/10/12
to Anush Elangovan(அனுஷ்), Chromium OS dev
This isn't working, tegra is still broken, and probably will be for a while because of missing hardfp support in various drivers.

Daisy, on the other hand, should now be working.

--
Chromium OS Developers mailing list: chromiu...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-os-dev?hl=en



--
Don

Mike Frysinger

unread,
May 10, 2012, 4:48:23 PM5/10/12
to Don Garrett, Anush Elangovan(அனுஷ்), Chromium OS dev
marcheu is working on adding hardfp drivers for tegra
-mike

Micah Catlin

unread,
May 10, 2012, 4:49:48 PM5/10/12
to Don Garrett, Anush Elangovan(அனுஷ்), Chromium OS dev, Stéphane Marchesin
Ideally we would have tested the tegra2 build at ToT with this
toolchain before changing the default.

But at the moment, we're in a situation where version 0.26 of the
various binary components (such as
opengles-bin-0.0.26.ebuild) don't respond to the "USE=hardfp" flag.
The previous versions (0.24) did support it, which is probably where
the toolchain team was doing their testing.

Since we can't roll back to 0.24 (due to 0.26 being the first which
support X abi 12) we're going to have to go forward with new ebuilds
that select the hardfp binaries. Nvidia has (just this week) posted
some release-candidate quality drivers built for hardfp, so there's a
chance of getting this done quickly.

Anush Elangovan(அனுஷ்)

unread,
May 10, 2012, 4:55:06 PM5/10/12
to Micah Catlin, Don Garrett, Stéphane Marchesin, Chromium OS dev


On May 10, 2012 1:49 PM, "Micah Catlin" <mic...@chromium.org> wrote:
>
> Ideally we would have tested the tegra2 build at ToT with this
> toolchain before changing the default.
>
> But at the moment, we're in a situation where version 0.26 of the
> various binary components (such as
> opengles-bin-0.0.26.ebuild) don't respond to the "USE=hardfp" flag.
> The previous versions (0.24) did support it, which is probably where
> the toolchain team was doing their testing.
>
> Since we can't roll back to 0.24 (due to 0.26 being the first which
> support X abi 12) we're going to have to go forward with new ebuilds
> that select the hardfp binaries.  Nvidia has (just this week) posted
> some release-candidate quality drivers built for hardfp, so there's a
> chance of getting this done quickly

Why dont we mark the tegra2 builders not important right now?  And just get it enough to build.

.

Peter Mayo

unread,
May 10, 2012, 4:58:38 PM5/10/12
to Anush Elangovan(அனுஷ்), Micah Catlin, Don Garrett, Stéphane Marchesin, Chromium OS dev
Why don't we just remove them if they don't matter?

Peter Mayo | Waterloo |  pete...@google.com |  519-880-3439


Zelidrag Hornung

unread,
May 10, 2012, 4:59:17 PM5/10/12
to Anush Elangovan(அனுஷ்), Micah Catlin, Don Garrett, Stéphane Marchesin, Chromium OS dev
Anush,

Why have you guys done this toolkit change without assuring that everything really worked first?

Don Garrett

unread,
May 10, 2012, 5:00:06 PM5/10/12
to Peter Mayo, Anush Elangovan(அனுஷ்), Micah Catlin, Stéphane Marchesin, Chromium OS dev
Both arm and daisy are already marked as unimportant, but the tree is still throttled since things are in a fragile/broken state that's already complex and hard to fix.
--
Don

Anush Elangovan(அனுஷ்)

unread,
May 10, 2012, 5:01:08 PM5/10/12
to Micah Catlin, Don Garrett, Stéphane Marchesin, Chromium OS dev


On May 10, 2012 1:55 PM, "Anush Elangovan(அனுஷ்)" <an...@chromium.org> wrote:
>
>
> On May 10, 2012 1:49 PM, "Micah Catlin" <mic...@chromium.org> wrote:
> >
> > Ideally we would have tested the tegra2 build at ToT with this
> > toolchain before changing the default.
> >
> > But at the moment, we're in a situation where version 0.26 of the
> > various binary components (such as
> > opengles-bin-0.0.26.ebuild) don't respond to the "USE=hardfp" flag.
> > The previous versions (0.24) did support it, which is probably where
> > the toolchain team was doing their testing.
> >
> > Since we can't roll back to 0.24 (due to 0.26 being the first which
> > support X abi 12) we're going to have to go forward with new ebuilds
> > that select the hardfp binaries.  Nvidia has (just this week) posted
> > some release-candidate quality drivers built for hardfp, so there's a
> > chance of getting this done quickly
>
> Why dont we mark the tegra2 builders not important right now?  And just get it enough to build.

FYI: this is a run time failure which is being addressed in parallel to getting the tree in shape.

Mike Frysinger

unread,
May 10, 2012, 5:03:22 PM5/10/12
to Zelidrag Hornung, Anush Elangovan(அனுஷ்), Micah Catlin, Don Garrett, Stéphane Marchesin, Chromium OS dev
this isn't really helping. please focus on that.
-mike

Anush Elangovan(அனுஷ்)

unread,
May 10, 2012, 5:08:32 PM5/10/12
to Zelidrag Hornung, Don Garrett, Stéphane Marchesin, Chromium OS dev, Micah Catlin


On May 10, 2012 1:59 PM, "Zelidrag Hornung" <zeli...@chromium.org> wrote:
>
> Anush,
>
> Why have you guys done this toolkit change without assuring that everything really worked first?

There were specific instructions given to rolling out the tool chain (i.e involving build team to roll it out properly ). However these were ignored and the cl landed before the infrastructure was prepped up, leading to a domino effect of trying to revive the situation.

Tegra2 is a best effort right now since it is actually forked already and is dragging us behind. But we hope it will be back in some working form in parallel

Anush Elangovan(அனுஷ்)

unread,
May 10, 2012, 5:34:13 PM5/10/12
to Zelidrag Hornung, Don Garrett, Stéphane Marchesin, Chromium OS dev, Micah Catlin
FYI. The tree is now open. tegra2 runtime will be fixed in parallel.

Thanks for everyone who helped revive the situation.


On Thu, May 10, 2012 at 2:08 PM, Anush Elangovan(அனுஷ்)

Don Garrett

unread,
May 10, 2012, 5:35:27 PM5/10/12
to Anush Elangovan(அனுஷ்), Zelidrag Hornung, Stéphane Marchesin, Chromium OS dev, Micah Catlin
I would like to have a post mortem after tegra2 is fixed.
--
Don

Chris Sosa

unread,
May 10, 2012, 5:37:43 PM5/10/12
to Don Garrett, Anush Elangovan(அனுஷ்), Zelidrag Hornung, Stéphane Marchesin, Chromium OS dev, Micah Catlin
Agreed. It would especially like to see if we could file any tasks on
the build team to make testing such migrations easier.

-Sosa

Anush Elangovan(அனுஷ்)

unread,
May 10, 2012, 5:40:00 PM5/10/12
to Chris Sosa, Don Garrett, Zelidrag Hornung, Stéphane Marchesin, Chromium OS dev, Micah Catlin
On Thu, May 10, 2012 at 2:37 PM, Chris Sosa <so...@chromium.org> wrote:
> Agreed.  It would especially like to see if we could file any tasks on
> the build team to make testing such migrations easier.

http://code.google.com/p/chromium-os/issues/detail?id=26231

Toolchain team has been asking for a way to gate their changes with a
CQ (ish) thing. If we can get that it would be great.

Thanks
Anush

Chris Masone

unread,
May 10, 2012, 6:00:15 PM5/10/12
to Anush Elangovan(அனுஷ்), Chris Sosa, Don Garrett, Zelidrag Hornung, Stéphane Marchesin, Chromium OS dev, Micah Catlin
On Thu, May 10, 2012 at 2:40 PM, Anush Elangovan(அனுஷ்) <an...@chromium.org> wrote:
On Thu, May 10, 2012 at 2:37 PM, Chris Sosa <so...@chromium.org> wrote:
> Agreed.  It would especially like to see if we could file any tasks on
> the build team to make testing such migrations easier.

http://code.google.com/p/chromium-os/issues/detail?id=26231

Toolchain team has been asking for a way to gate their changes with a
CQ (ish) thing. If we can get that it would be great.


While it's clear that such a thing would help, I think a lot of folks would benefit from seeing a real postmortem of what happened this morning.
Reply all
Reply to author
Forward
This conversation is locked
You cannot reply and perform actions on locked conversations.
0 new messages