Crash in HTTP2Transport::cleanupFinishedStreams()

[garmin-holder]

We are getting several crashes in two different place in cleanupFinishedStreams (line 354 and 393) in 0.3. We upgraded to 0.3 in hopes that the issue was fixed, but we are still getting it. Is this a known issue and if so is it fixed on 0.4? Interestingly we only seem to be getting in in iOS, not Android.

Crashed: Thread
0 Gecko 0x1006eed28 alexaClientSDK::acl::HTTP2Transport::cleanupFinishedStreams() (HTTP2Transport.cpp:354)
1 Gecko 0x1006ee5f4 alexaClientSDK::acl::HTTP2Transport::networkLoop() (HTTP2Transport.cpp:265)
2 Gecko 0x1006ef738 void* std::__1::__thread_proxy<std::__1::tuple<void (alexaClientSDK::acl::HTTP2Transport::)(), alexaClientSDK::acl::HTTP2Transport> >(void*) (memory:2525)
3 libsystem_pthread.dylib 0x189ca568c _pthread_body + 240
4 libsystem_pthread.dylib 0x189ca559c _pthread_body + 282
5 libsystem_pthread.dylib 0x189ca2cb4 thread_start + 4

[scotthea-amazon]

Hello,

There is fix to cleanupFinishedStreams() in the 0.4 release for a segmentation fault. In the 0.4 source you will see that line 391:

curl_multi_remove_handle(m_multi->handle, message->easy_handle);

has been moved from just after line 393. This corrected a segmentation fault that we were seeing occasionally in the ACL integration test. It was caused by a mis-ordring tear-down of libcurl resources. This fix should be appropriate in the 0.3 source as well.

I am not 100% sure that this is the same as the issue you are encountering. The lines numbers in the stack trace you provided do not both match up with cleanupFinishedStreams() in the 0.2 and 0.3 source I am looking at. Perhaps you could provide the source file and I could take a closer look.

Please let us know if this helps,

-SWH

[garmin-holder]

That is probably the same issue. I believe the 354 crash was with 0.2 and 393 reflects an extra include we had to add in 0.3 for Android. Thanks for the response.

[scotthea-amazon]

A fix for this has been pushed in version 0.4.1.