New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 607312 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Last visit > 30 days ago
Closed: May 2016
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug

Blocking:
issue 595702
issue 606810



Sign in to add a comment

Roll Clang

Project Member Reported by h...@chromium.org, Apr 27 2016

Issue description

krasin wants r267784

and we need r267634 for  crbug.com/606810 
 

Comment 1 by krasin@chromium.org, Apr 28 2016

We also need r267824: http://reviews.llvm.org/rL267824
This should fix 'CFI Linux ToT' buildbot:  https://crbug.com/607370 

Comment 2 by krasin@chromium.org, Apr 28 2016

Ah, I said a stupid thing. Please, read it as "we need to roll a toolchain past r267824 to avoid breaking 'CFI Linux' buildbot'"

Comment 3 by krasin@chromium.org, Apr 29 2016

Cc: kcc@chromium.org p...@chromium.org
Hans, Nico,

it seems that we (me and Peter) are the primary customers for the upcoming roll. Is there anything I can do to help with that? Theoretically, picking a good revision, running the script, waiting for clang upload trybots and then regular trybots to succeed is not super complicated

Do you do any additional testing while rolling new Clang?

Comment 4 by h...@chromium.org, Apr 29 2016

> Theoretically, picking a good revision, running the script, waiting for clang upload trybots and then regular trybots to succeed is not super complicated

Yup, it should be pretty easy now :-)

Except, we run some more try jobs besides the usual ones, see the list at https://chromium.googlesource.com/chromium/src/+/master/docs/updating_clang.md

And in between upload bots and running tryjobs, the binary needs to be pushed to goma.

Do you want to kick off a build now, or should we wait until Monday?

Comment 5 by thakis@chromium.org, Apr 29 2016

Oh, I had the impression that the revision you were waiting for was reverted upstream. Ah, it was, and I had missed the reland in r267903.

We mostly check that https://build.chromium.org/p/chromium.fyi/console?category=clang%20tot is mostly green. Looks like clang's currently asserting when building chrome/ios: https://build.chromium.org/p/chromium.fyi/builders/ClangToTiOS/builds/6624/steps/compile/logs/stdio Things were looking pretty good a few builds earlier though, so maybe you can pick a revision slightly older than trunk.

Thanks to your work, it is in fact pretty easy to build new packages: Just pick a revision, run your script (ideally in a chromium checkout that's synced up, so that you pick up recent plugin changes), and once it's done ask me or Hans to upload to goma. Then wait for it to replicate to enough slaves, then send many try jobs (list here: https://chromium.googlesource.com/chromium/src/+/master/docs/updating_clang.md -- the rest of that page looks pretty out of date though :-( Maybe your script could have a "send try jobs" command too), and if those are green, hit cq.

Comment 6 by krasin@chromium.org, Apr 29 2016

Monday is probably better. Thank you for the instructions, assigning the issue to myself. I may need additional guidance from you, as it's the first time I roll the toolchain, but overall the procedure is pretty clear.

To be honest, the doc ([1]) needs to be updated as it does not mention clang upload trybots. I will fix it when the roll is done.

1. https://chromium.googlesource.com/chromium/src/+/master/docs/updating_clang.md

Comment 7 by krasin@chromium.org, Apr 29 2016

Owner: krasin@chromium.org
Blocking: 595702
We want r268331 too
As a quick non-update: I plan to start working on the roll tomorrow. This morning the ToT weather didn't look great for Windows, so I decided to spend time on something else.
I kicked off an upload run this morning, it's currently running here: https://codereview.chromium.org/1944713002/

All the clang/win bots are currently red due to  bug 608705  (the pinned ones too) though.
I triggered all the trybots listed in https://chromium.googlesource.com/chromium/src/+/master/docs/updating_clang.md + CQ dry run.
Cc: euge...@chromium.org
So, the try bots are mostly green, except of one msan bot:
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_msan_rel_ng/builds/337

I am currently building it locally to reproduce and fix (if it's a real report)
I could never reproduce it locally. If you can, let's fix it, probably with an attribute or an ifdef.
I think msan always comes back red :-/
Shall we go ahead?

FWIW, I've reproduced the bug locally, and will spend some time with Eugene trying to fix it. Given that it's always red, that does not sound like a blocker for the roll.
If you lgtm the roll CL, I can hit cq+ :-)
Done :)
Project Member

Comment 19 by bugdroid1@chromium.org, May 4 2016

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/79e7ee8785d99faf38115bdc08fcb1ada0239166

commit 79e7ee8785d99faf38115bdc08fcb1ada0239166
Author: thakis <thakis@chromium.org>
Date: Wed May 04 03:14:30 2016

Roll clang 267383:268373.

Ran `tools/clang/scripts/upload_revision.py 268373`.

BUG= 607312 

Review-Url: https://codereview.chromium.org/1944713002
Cr-Commit-Position: refs/heads/master@{#391432}

[modify] https://crrev.com/79e7ee8785d99faf38115bdc08fcb1ada0239166/tools/clang/scripts/update.py

Status: Fixed (was: Available)

Sign in to add a comment