[lldb-dev] lldb hangs during reverse-connect if lldb-server dies

[Greg Clayton]

We should implement a timeout of say 5 seconds (since this is a locally spawned lldb-server) and also watch for the spawned lldb-server process to be reaped. If the lldb-server gets reaped we need to cancel. Don't do any fancy thread canceling (no pthread_cancel or anything like that) if you do fix this, just have something locally connect to the socket and allow the accept thread to exit normally.


> On Jun 17, 2015, at 11:32 AM, Ted Woodward <ted.woodward at codeaurora.org> wrote:
> 
>  
> With reverse-connect, lldb gets a port from the OS, spawns a listen thread using the port, and launches lldb-server/debugserver, telling it to connect back to lldb. lldb with then join with the listen thread. The listen thread calls accept() to wait for the server to connect back to it.
>  
> On Linux (and possibly other OSes), if lldb-server dies or quits before connecting, lldb will hang. It sits waiting for accept() to return, but it never will.
>  
> If the child dies, the listen thread should be killed, and an error returned to the user.
>  
>  
> This is easily seen on Linux. Simply point to an alternate debugserver. Set LLDB_DEBUGSERVER_PATH to something else, say /bin/ls. This will finish and never connect back to lldb.
>  
> -> setenv LLDB_DEBUGSERVER_PATH /bin/ls
> ->bin/lldb /bin/ls
> (lldb) target create "/bin/ls"
> Current executable set to '/bin/ls' (x86_64).
> (lldb) r
>  
> <hang>
>  
> --
> Qualcomm Innovation Center, Inc.
> The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
>  
> _______________________________________________
> lldb-dev mailing list
> lldb-dev at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev