Wrapper around pdb allowing remote debugging via netcat or telnet. This is especially useful in a Tomcat/Jython environment where little debugging tools are available.
BSD 2-Clause "Simplified" License
130
stars
44
forks
source link
fix regressions triggered by (c)ontinue calls during a session #30
There were also a couple other issues...
First and foremost, this resolves the regression described in issue #29 caused by commit 18cd5a4. We want to allow the debugger to continue, but we don't want to close the connection on the socket until it's actually time to shut down (EOF or SIGQUIT or an exception occurs etc).
Secondly, the replacement of the sys.stdin and sys.stdout as file objects is not enough - we want to update the underlying file descriptor because there may be some other underlying object that saved/uses the old stdin/stdout. As such, to we use dup2 to ensure subsequent reads/writes go into the socket. Though honestly, it's a bit redundant because the PDB/BDB class inherits the Cmd class which reads from the socket implicitly anyway (passing the socket handle as stdin/stdout is sufficient).
Anyway, this should at least make it functional again.
There were also a couple other issues... First and foremost, this resolves the regression described in issue #29 caused by commit 18cd5a4. We want to allow the debugger to continue, but we don't want to close the connection on the socket until it's actually time to shut down (EOF or SIGQUIT or an exception occurs etc).
Secondly, the replacement of the sys.stdin and sys.stdout as file objects is not enough - we want to update the underlying file descriptor because there may be some other underlying object that saved/uses the old stdin/stdout. As such, to we use dup2 to ensure subsequent reads/writes go into the socket. Though honestly, it's a bit redundant because the PDB/BDB class inherits the Cmd class which reads from the socket implicitly anyway (passing the socket handle as stdin/stdout is sufficient).
Anyway, this should at least make it functional again.