Commits

Show all
Author Commit Message Labels Comments Date
Charlie Groves
Return Py.None instead of null from builtin methods like execfile that don't return anything. This fixes the root cause of bug #1758838, so I'm rolling back r3355 since that fix is no longer needed.
Tags
v2.2rc3
Charlie Groves
Note latest bugs fixed
Charlie Groves
Fix the test on both input and output so it attempts to read the amount written from the client on the server, and the amount written on the server from the client instead of vice versa. Restructure the tests as a more traditional set of unit tests on client writing, server writing and simultaneous writing instead of one long set of dependent tests.
Charlie Groves
Set SO_REUSEADDR in setting up client connections for Threaded[TCP|UDP] tests. Break out the server only portions of NonBlockingTCPTests into their own class so we're not firing up threads to test them with the client side needlessly.
Charlie Groves
Move the call to _setreuseaddress into _do_connect so it always takes effect before bind is called on the java socket.
Charlie Groves
Don't attempt to ensure that the socket isn't writable after fill_outchannel since the vagaries of buffering means that may not always be the case.
Charlie Groves
Only select on the direction we're testing: in or out. Java's select seems to do some oddball things if the availability of one direction has just changed and you express interest in both.
Charlie Groves
enable REUSEADDR in TestJythonExceptions so the initial bind in testBindException will succeed. Switch TCPClientTimeoutTest from trying to create a partially setup server socket and then timing out connecting to it to connecting to a (hopefully) unused external IP.
Charlie Groves
Set reuse_addr on _tcpsocket wrappers to the value of getReuseAddress on the Java Socket instance it's going to wrap.
Oti Humbel
fix and bugtest for bug 1758838 (NullPointerException in execfile(), in interactive mode)
Charlie Groves
Wait for self.done before returning from testTCPClientTimeout so the server socket isn't closed before the client has a chance to timeout on it on multi-processor machines. This doesn't fix the failure on my mac, but it at least keeps the socket open long enough for it to fail to timeout instead of it not even being open when it tries. Always explicitly close the socket in TestJythonExceptions. The socket wasn't being garbage collected quickly enough to close it so it wasn't able to bind again in testBindException.
Charlie Groves
test_socket uses unittest for its testing, so don't include any output for comparison
Charlie Groves
Use test_support to run tests so regrtest picks them up
Charlie Groves
Version bumpage for 2.2rc3
Charlie Groves
telnetlib is in Lib now, it doesn't need to be included
Charlie Groves
Interpret -1 to mean the socket was closed, and return the empty string from recv in that case to match CPython.
Charlie Groves
Use the regular file close mechanism if we haven't created a closer since this PyFile was created as a Java class. Always use a closer if the PyFile is being created as a builtin.
Charlie Groves
use cpython_compatible_select in place of select to fix the last of bug #1755096.
Charlie Groves
Alan Kennedy
. Added a socket.getblocking() method, and unittest for same . Tidied up select exception handling . Removed ERRNO_ constants from select module; jython 2.1 users can create their own errno module. . Some small speed improvements to select.select . Added a cpython compatible select function to select module
Charlie Groves
only automatically close PyFiles if they're created with a string filename from Python
Alan Kennedy
. Fixed a bug with FileWrappers, whereby closing the FileWrapper incorrectly caused closing of the underlying socket. . Wrapped several more methods in try .. except for exception mapping . Removed unnecessary lambdas from exception mapping table . Completed the exception mapping table with all java.net and java.nio exceptions . Removed errno module simulation: was unnecessary (jython 2.1 users can easily create their own errno module) . Added constants SHUT_RD, SHUT_WR and SHUT_RDWR, as used by shutdown() . Adde…
Charlie Groves
Add remaining bugs fixed in 2.2rc2
Charlie Groves
Making 2.2rc2 tag.
Charlie Groves
Version bump for 2.2rc2
Tags
v2.2rc2
Charlie Groves
The last commit didn't actually fix bug #1747092. It got a somewhat similar but different bug where importing a module that throws an exception on import added it to sys.modules. That caused it to be silently imported as an empty module without an exception if tried again. This commit keeps top-level imports from being added to sys.modules if they're not found, so if sys.path is corrected and the import is tried again it'll actually work. It still adds None for subpackage and submodule imports so they're not tried again after they fail.
Charlie Groves
Remove a module from sys.modules if it blows up while being imported. Fixes bug #1747092
Charlie Groves
Don't add a u prefix while unicode escaping a value > 255 if working on a unicode object since it always has the u. Fixes bug #1746957
Charlie Groves
Fix for bug #1719528 where -c didn't add the empty string to sys.path
Charlie Groves
Patch #1746339 from mehendran to fix bug #931129 where a jython jar using __run__.py couldn't have certain directory structures in its path
  1. Prev
  2. Next