Murphy: notice: warning: system call 270 (pselect6) not supported for program

[karlp]

Was trying this out after reading regehr's blog, thought it looked interesting.

Is this actually still in development at all? has it moved on into another project? Running this on my somewhat simple but admittedly threaded code didn't reall do anything interesting....

[karlp]

A second application produced no warnings, but failed to operate, Murphy seemed to interfere with opening a TCP socket as well, I get from strerror, "Address family not supported by protocol"

[liblit]

Murphy is not under active development, but we certainly want it to remain working and available for anyone interested in using or extending it. Thanks for giving it a try!

Regarding “threaded code”, I don’t think Murphy has any problem with threads. But I will double-check with Zach Miller and Todd Tannenbaum, who did the real development work.

Regarding “didn’t really do anything interesting”, what gremlin configuration are you using? The standard murphy_config from the top of the source tree only has two gremlins enabled. You might try enabling a few more as shown in murphy_config.example to see if that triggers any interesting failures. You can also run Murphy with the -d murphy flag to get extra debug output. That would at least confirm for you that Murphy is indeed active.

Regarding “system call 270 (pselect6) not supported for program”, that's definitely coming out of the parrot library that Murphy uses for system call interception. My guess would be that this is a newer system call for which parrot simply does not have a handler. I am not familiar with the parrot code myself, so I am not sure what changes would be appropriate to add pselect6 system call handling. I will pass this along to Zach and Todd to see if they have any suggestions

Regarding “Address family not supported by protocol”, I really have no way of knowing what is going on there without more information. Perhaps this is a defect in our system call handling; perhaps it is a latent defect in your application code that Murphy is exposing. I can try to explore that further if you provide code that reproduces the bug. Please file that as a distinct github issue, though, unless there is some reason to believe it is tied to the pselect6 problem you initially reported here.