Join GitHub today
GitHub is home to over 20 million developers working together to host and review code, manage projects, and build software together.
fix issues caught by -fsanitize=bounds in bluetooth.default #436
Comments
jgeerds
referenced this issue
Sep 17, 2016
Closed
exec spawning loses capabilities, breaking bluetooth and wakelocks #398
thestinger
added
Type: bug
upstream
labels
Sep 17, 2016
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
thestinger
Sep 17, 2016
Contributor
This would be an upstream bug caught by -fsanitize=bounds. This will have to be yet another library given an exception from this.
|
This would be an upstream bug caught by -fsanitize=bounds. This will have to be yet another library given an exception from this. |
thestinger
added
the
Status: workaround in place
label
Sep 17, 2016
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
jgeerds
Sep 17, 2016
So the proper solution would be to wait until it's fixed in upstream or is there another way? Do you have a quick workaround for it? (like disabling the bluetooth service)
jgeerds
commented
Sep 17, 2016
|
So the proper solution would be to wait until it's fixed in upstream or is there another way? Do you have a quick workaround for it? (like disabling the bluetooth service) |
thestinger
changed the title from
Bluetooth constantly crashing since latest update
to
fix issues caught by -fsanitize=bounds in bluetooth.default
Sep 17, 2016
thestinger
added
the
Component: ubsan
label
Sep 17, 2016
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
thestinger
Sep 17, 2016
Contributor
It has already been worked around by disabling -fsanitize=bounds for it. The issue has to remain open until the memory corruption bug(s) in the bluetooth library are fixed. You can probably avoid the crashes by disabling bluetooth but this specific issue will already be worked around in the next release.
|
It has already been worked around by disabling -fsanitize=bounds for it. The issue has to remain open until the memory corruption bug(s) in the bluetooth library are fixed. You can probably avoid the crashes by disabling bluetooth but this specific issue will already be worked around in the next release. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
jgeerds
Sep 17, 2016
For some reason it is not possible to disable Bluetooth from the Android GUI. The phone immediately tries to activate bluetooth again.
So there will be a new release in the next few hours/days? Is there a place where I can monitor new releases (like your build pipeline)?
jgeerds
commented
Sep 17, 2016
|
For some reason it is not possible to disable Bluetooth from the Android GUI. The phone immediately tries to activate bluetooth again. So there will be a new release in the next few hours/days? Is there a place where I can monitor new releases (like your build pipeline)? |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
thestinger
Sep 17, 2016
Contributor
So there will be a new release in the next few hours/days?
Not in a few hours. It takes ~2-3 hours to build + sign + upload for each device and more time for testing. Perhaps today though.
Is there a place where I can monitor new releases (like your build pipeline)?
No. All you can do is see that it's there once it's published.
Not in a few hours. It takes ~2-3 hours to build + sign + upload for each device and more time for testing. Perhaps today though.
No. All you can do is see that it's there once it's published. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
jgeerds
Sep 17, 2016
btw. I just donated $25 to you/copperhead. Thanks for you fast support!
jgeerds
commented
Sep 17, 2016
|
btw. I just donated $25 to you/copperhead. Thanks for you fast support! |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment|
Thanks! |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
subproc
commented
Sep 17, 2016
|
nexus 6p too is in a crash loop |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment|
I know it's not a device-specific issue. It's fixed already. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
KnucklesPierce
Sep 17, 2016
Hello Stinger, having the same issue here; just updated this morning to NRD90U.2016.9.16.14.31.32. I was able to shut off BT to stop getting the errors.
You mentioned it is fixed already...am I not running the latest build?
Is there a way to reinstall the latest build to rule out possibility of a bad install?
Thanks for your help
KnucklesPierce
commented
Sep 17, 2016
|
Hello Stinger, having the same issue here; just updated this morning to NRD90U.2016.9.16.14.31.32. I was able to shut off BT to stop getting the errors. You mentioned it is fixed already...am I not running the latest build? Is there a way to reinstall the latest build to rule out possibility of a bad install? Thanks for your help |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment|
It isn't released yet. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
KnucklesPierce
commented
Sep 17, 2016
|
Ah, gotcha. Am I able to revert back to a previous build? |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
No. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
KnucklesPierce
commented
Sep 17, 2016
|
Damn. Thanks for your time and quick replies, Dan. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
thestinger
Sep 17, 2016
Contributor
The new builds would already be published if we had a proper build server, and problems like this could be avoided in the first place with enough devices to make running the full CTS for every release reasonable by spreading it out. There is going to be a LOT more of this for -fsanitize=bounds once it's enabled for C++ again rather than just C. This is one of the painful parts of migrating to Android Nougat.
|
The new builds would already be published if we had a proper build server, and problems like this could be avoided in the first place with enough devices to make running the full CTS for every release reasonable by spreading it out. There is going to be a LOT more of this for -fsanitize=bounds once it's enabled for C++ again rather than just C. This is one of the painful parts of migrating to Android Nougat. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
KnucklesPierce
Sep 17, 2016
I gathered so much from reading some of the posts James has been sending out lately. I'm not a programmer by any stretch of the imagination (last program I made was in VB5), so I'm not completely sure what I read some of the time, but the underlying tones of frustration come through nonetheless. I actually made a mental note to not upgrade due to those posts...that went out the window at 5am.
On a semi-related topic, what would be a proper build server?
KnucklesPierce
commented
Sep 17, 2016
|
I gathered so much from reading some of the posts James has been sending out lately. I'm not a programmer by any stretch of the imagination (last program I made was in VB5), so I'm not completely sure what I read some of the time, but the underlying tones of frustration come through nonetheless. I actually made a mental note to not upgrade due to those posts...that went out the window at 5am. On a semi-related topic, what would be a proper build server? |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
thestinger
Sep 17, 2016
Contributor
The message that was meant to be communicated was not that you shouldn't upgrade. You need to stay updated to get the security updates. There was an initial issue with the Updater app but it was known how to work around that before any announcement about it was made. The migration is already done and it's back to how things are regularly.
|
The message that was meant to be communicated was not that you shouldn't upgrade. You need to stay updated to get the security updates. There was an initial issue with the Updater app but it was known how to work around that before any announcement about it was made. The migration is already done and it's back to how things are regularly. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
thestinger
Sep 17, 2016
Contributor
On a semi-related topic, what would be a proper build server?
A dual socket server with 16 core Xeon E5 CPUs at a minimum. Ideally much better than that. Builds take a ridiculous amount of time and it's a huge bottleneck on the development process. Doing more than basic QA to move the project out of beta requires having someone else working on the project full-time and a lot of phones for running the CTS in a reasonable time period. The project will be eternally in beta until there are several more people working on it full-time.
A dual socket server with 16 core Xeon E5 CPUs at a minimum. Ideally much better than that. Builds take a ridiculous amount of time and it's a huge bottleneck on the development process. Doing more than basic QA to move the project out of beta requires having someone else working on the project full-time and a lot of phones for running the CTS in a reasonable time period. The project will be eternally in beta until there are several more people working on it full-time. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
thestinger
Sep 17, 2016
Contributor
CopperheadOS implements new mitigations and they end up discovering upstream Android bugs. It's working as intended. There will be more of these issues found when -fsanitize=bounds it's enabled for C++ again rather than only C. For these I only need an one report of how to trigger the issue with a log and then it can be worked around by disabling the feature for that broken code until it's fixed.
Only issues breaking the booting / update process or network access are going to be discovered by the current QA done for each release. Anything else has to be reported. A subset of the CTS is occasionally run as part of the development process, but not for every release, and rarely the entire thing.
|
CopperheadOS implements new mitigations and they end up discovering upstream Android bugs. It's working as intended. There will be more of these issues found when -fsanitize=bounds it's enabled for C++ again rather than only C. For these I only need an one report of how to trigger the issue with a log and then it can be worked around by disabling the feature for that broken code until it's fixed. Only issues breaking the booting / update process or network access are going to be discovered by the current QA done for each release. Anything else has to be reported. A subset of the CTS is occasionally run as part of the development process, but not for every release, and rarely the entire thing. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment|
The new build is released. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
thestinger
Sep 18, 2016
Contributor
Closing in favour of #439 which tracks fixing the upstream memory corruption bug(s).
|
Closing in favour of #439 which tracks fixing the upstream memory corruption bug(s). |
thestinger
closed this
Sep 18, 2016
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
Show comment Hide comment
jgeerds
commented
Sep 18, 2016
|
I can confirm that the new build works on my Nexus 5X. Thank you very much! |
jgeerds commentedSep 17, 2016
My Bluetooth is constantly crashing since I updated to the latest version (
2016-09-16 14:31:32) on my Nexus 5X. The previous versin from 2016-09-14 worked fine for me.Here is the shortened output of
logcat. I think it covers the important parts (it is basically a loop of this error)