Segfaults on connect from Mac to Ubuntu about eternalterminal

[yixizhang]

Same here.

[MisterTea]

It may be the double-free issue. Can you get me a stack trace? You'll have to edit the et launcher to run etclient in lldb / gdb.

…

On Sun, Jun 25, 2017 at 11:17 AM Yixi Zhang ***@***.***> wrote: Same here. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#43 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AASuv5d8g3FjPPvgjDmaJPkIzvLOMgKEks5sHqQsgaJpZM4OEhbY> .

[MisterTea]

change this line:

CLIENT_BINARY="etclient"

to this:

CLIENT_BINARY="lldb -- etclient"

then rerun with lldb and when it crashes, type "bt" to give me the stack trace

[gramster]

(lldb) r
Process 2007 launched: '/usr/local/bin/etclient' (x86_64)
Process 2007 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
    frame #0: 0x00007fffc8fdcb52 libsystem_c.dylib`strlen + 18
libsystem_c.dylib`strlen:
->  0x7fffc8fdcb52 <+18>: pcmpeqb (%rdi), %xmm0
    0x7fffc8fdcb56 <+22>: pmovmskb %xmm0, %esi
    0x7fffc8fdcb5a <+26>: andq   $0xf, %rcx
    0x7fffc8fdcb5e <+30>: orq    $-0x1, %rax
(lldb) bt
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
  * frame #0: 0x00007fffc8fdcb52 libsystem_c.dylib`strlen + 18
    frame #1: 0x0000000100011e77 
etclient`et::UnixSocketHandler::connect(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, int) + 2355
    frame #2: 0x000000010000da32 etclient`et::ClientConnection::connect() + 272
    frame #3: 0x00000001000038fd etclient`createClient() + 2375
    frame #4: 0x0000000100004579 etclient`main + 147
    frame #5: 0x00007fffc8fa6235 libdyld.dylib`start + 1

[MisterTea]

Are you running the latest homebrew version? Those addresses don't line up for me:

(lldb) target create --no-dependents --arch x86_64 /usr/local/bin/etclient
Current executable set to '/usr/local/bin/etclient' (x86_64).
(lldb) image list
[  0] C6084147-6006-33B2-AEC7-3E411FECE663 0x0000000100000000 /usr/local/bin/etclient 
(lldb) image lookup --address 0x0000000100011e77
      Address: etclient[0x0000000100011e77] (etclient.__TEXT.__text + 62311)
      Summary: etclient`et::UnixSocketHandler::connect(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, int) + 1407

[MisterTea]

Either way I think I found the problem. Going to patch now.

[MisterTea]

@gramster Please build the latest master and verify that the problem is fixed, thanks!

[MisterTea]

BTW: This is a client only bug so you don't have to change the server binary

[yixizhang]

built master on OSX, no more above error but can't connect to host.

Could not make initial connection to $ip: Could not connect to host

[yixizhang]

Validated that ssh with same user@ip works.

[MisterTea]

@yixizhang Are you using ssh_config aliases? These aren't supported yet. If not, can you post the whole command in context?

[yixizhang]

Sorry, I forgot that I used ssh-agent and ssh-add when I validated ssh and et yesterday.

The command I used for ssh is ssh -i .ssh/google_compute_engine user@host. I tried to use the corresponding et command, in format of et -s "-i .ssh/google_compute_engine user@host:port" but it complains format:

Usage: ET (options) [user@]hostname[:port]

[CallumDenby]

@MisterTea I can confirm that building the latest version on Mac fixes the segmentation fault. All is working for me now.

macOS 10.12.5, et initially installed with brew, then etclient built from source and original overwritten.

[MisterTea]

Fixed in https://github.com/MisterTea/EternalTCP/releases/tag/et-v3.1.0