fix: work around a null scoped_ptr dereference #19366
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
CI seems to have flaked due the Github downtime. |
|
Have you observed the log message being printed at all? I'm not sure what would cause the callback to become nulled out... |
|
So I just realised that I forgot to add |
|
OK, so a quick update on this one: I haven't been able to find an acceptable way of adding |
|
OK, so still trying to track down the problem. Everything seems to be behaving a bit weirdly, and my fix is not actually effective (?) Here's my x64 system demonstrating it: So 1) The callback's |
|
@richard-townsend-arm to see the logging set the ELECTRON_ENABLE_LOGGING environment variable to 1. |
|
@jkleinsc EDIT: oh, i see, the earlier message was saying that passing the flag wasn't easy to do. carry on. |
|
So interestingly (trying with and without --ci), seems that it may be happening in the webview section: stderr is not very helpful still: |
|
I'll try to replicate it on Linux (that way I can do reverse debug). |
|
So I think the missing piece of the puzzle is that |
richard-townsend-arm
force-pushed
the
null-cb-deref
branch
from
fc4f1b2
to
f947ebc
Compare
|
Cool, so the current patch should take care of the problem. |
4 tasks
|
@richard-townsend-arm can you rebase this with the latest from 6-0-x? I just merged in the changes so that we can properly build WOA. |
This happens occasionally when running the test suite and indicates that the callback's been reset or the underlying reference has been released. To workaround, print a warning.
richard-townsend-arm
force-pushed
the
null-cb-deref
branch
from
f947ebc
to
53286ce
Compare
|
I still don't understand how |
jkleinsc
approved these changes
Aug 27, 2019
|
No Release Notes |
|
(Although it's just been merged, a quick explanation of what I think was going wrong for posterity): I think it's related to #18065: when AtomCertVerifier::OnDefaultVerificationDone gets called, it doesn't check whether AtomCertVerifier::verify_proc() is null or not. It's supposed to be set to something non-null by SetCertVerifyProcInIO, but this is is (for some reason) queued up to run asynchronously on the IO thread. I think this means that when the IO thread's congested, there's a small window of opportunity where requests can complete and verify_proc() is null. It's proving tough to prove beyond all doubt that this is what's actually happening though due to the heavy asynchronous...ness going on. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description of Change
For the last week or so, we've been running each Electron 6 commit through the unit tests multiple times. Sometimes, the unit test suite dies halfway through for no discernible reason. Executing under the debugger indicates that the callback passed to
base::BindRepeatingduring atom's certificate verification process sometimes has a nullbind_state, which causes the crash (this happens on x86 and on Windows on Arm). To workaround, this PR changes the callback to one which logs the result if this occurs. Since applying this fix, we haven't seen the issue in subsequent runs.The certificate verification logic has been rewritten in master (therefore this patch will not apply). The new code may suffer from the same problem, but we haven't spotted it yet because we don't have enough capacity to test master and 6-0-x to the same level.
CC @jkleinsc @MarshallOfSound
Checklist
npm testpassesRelease Notes
Notes: no-notes