PCSX2 hang on exit (Linux)...

[gregory]

Hum is mtvu hack enabled ?

[toxicshadow]

With you mentioning graphics drivers.

My driver is the Nvidia 358.16 driver which is the first driver to add support for xorg 1.18..... Coincidence??

---

Everything is at default settings.

I literally delete the old build and rebuild a new one (with dev enabled).
Open the program.
Accept the plugins
Close the program
Hang

[gregory]

Please test without mtvu speedhack.

[toxicshadow]

Just tried again. MTVU is unchecked yet the hang still occurs.

Food for thought. I tried building and running the application in my virtual machine. That closes down as expected with no hang.

This does suggest that it might be external to the PCSX2 application

[toxicshadow]

Additional:

My virtual machine is Fedora 23 (32bit)
My main machine is Fedora 23 (64bit)

[turtleli]

Could you test git revision fa487474d2e4a4c9bae706ee5ddbd07473a97280? If PCSX2 closes normally try 372db4b539772991ea8efc53358ac94a9771b1f2.

[toxicshadow]

Good Call!

fa487474d2e4a4c9bae706ee5ddbd07473a97280 appears to work perfectly

372db4b539772991ea8efc53358ac94a9771b1f2 causes the hang

[turtleli]

Ok, thanks for testing, I'll try and improve the cleanup logic so it doesn't hang.

[toxicshadow]

Ok so I have done some debugging of my own by putting data out on the console after each step of the the cleanup process using "puts() commands.

Code:

void LinuxPipeRedirection::Cleanup() noexcept
{
    
    puts("a");
    
    // Restore stdout/stderr file descriptor - mostly useful if the thread
    // fails to start, but then you have bigger issues to worry about.
    if (m_pipe_fd[1] != -1) {
        puts("b");
        if (m_pipe_fd[1] == fileno(m_stdstream)) {
            puts("c");
            // FIXME: Use lock for better termination.
            // The redirect thread is most likely waiting at read(). Send a
            // newline so it returns and the thread begins to terminate.
            fflush(m_stdstream);
            puts("d");
            m_thread.Cancel();     <---------****** Fails Here *****
            puts("e");

The console shows:

Releasing host memory maps for virtual systems...
a
b
c
d
**** hangs here ****

It appears to hang at the m_thread.Cancel(); call above.

Hope it helps

[gregory]

Actually it appears on the backtrace. Actually it could just be a print on the cancel path