LTO test causing compilation error


I’m working on a patch a patch adding alignment to Compilation fails during a test that has seemingly nothing in common with my additions, deep into the LLVM LTO pipeline (which I have no idea what it is) while the test case contains no global operation.

Can someone tell me what I can do?


These tests are entirely unrelated, it is likely that if you rebase and update your revision the failure will be gone.

I’m sorry, I’m not very good with git. What does rebase mean? Just make a git pull, check that everything recompiles, and then call arc again?

I’ve just tried to do this: git pull, followed by git config pull.rebase true and by arc diff HEAD~. The result (after some steps) is the following message:

You don't own revision D102254: "[TableGen] Make the NUL character
invalid in .td files, version 2". Normally, you should only update
revisions you own. You can "Commandeer" this revision from the web
interface if you want to become the owner.

Update this revision anyway? [y/N] N

I did not proceed with this, and I don’t know how to deal with such conflicts. Help needed.


I don’t know what is the state of your branch right now, git status and git log may help you figure out.
Also I don’t know if you initial git pull before git config pull.rebase true was a rebase or a merge.

Here is the output of git status:

~/github/llvm-project > git status 
On branch main
Your branch is ahead of 'origin/main' by 2 commits.
  (use "git push" to publish your local commits)

nothing to commit, working tree clean

And here is the output of git log (the first screen):

commit cd102f39a0cd9c198ea4180cf4e5e66c8c0c037c (HEAD -> main)
Merge: bff0d8c2ab93 709f2c7e14c8
Author: Dumitru Potop <>
Date:   Fri May 14 20:28:24 2021 +0200

    Merge branch 'main' of into main

commit 709f2c7e14c80ef08dbc619bbc6f2a21fb2ab52d (origin/main, origin/HEAD)
Author: Tim Northover <>
Date:   Fri May 14 19:21:54 2021 +0100

    SwiftAsync: remove duplicate instance in array. NFC.

commit 23c93c255549f440444b10d73d439bedc0d6822d
Author: Philip Reames <>
Date:   Fri May 14 10:57:59 2021 -0700

    Discount invariant instructions in full unrolling
    This patch updates the cost model for full unrolling to discount the cost of a loop invariant expression on all but one iteration. The reasoning here is that such an expression (as determined by SCEV) will be CSEd or DSEd once the loop is unrolled. Note that SCEVs reasoning will find things which could be invariant, not simply those outside the loop.
    Differential Revision:

commit dfacb8c8d432f20b2ccb832b34bfab52a45784bb
Author: River Riddle <>
Date:   Fri May 14 10:58:29 2021 -0700

    [mlir] Add missing dependence to TestDialect from TestTransforms
    This was accidentally dropped in D102456

commit 5aafcb2b440fb71a026e9c74101f272caecf08b0
Author: Michael Kruse <>
Date:   Thu May 13 17:09:55 2021 -0500

    [Polly] Add support for -polly-position=early with the NPM.

Yeah your initial git pull actually merged the main branch instead of rebasing. You may wanna try:

git reset --hard bff0d8c2ab93
git pull --rebase

(the --rebase may not be necessary if you have it in the config)

Then git log should show your commit on top of the updated branch.

1 Like

Ok, I would also really like to know how to send a message that does not get registered on the forum but somehow ends up in my mailbox as part of this discussion. It comes from none via LLVM Discussion Forums. Is this a security breach?

No, it is a message I deleted through moderation.