Updated README.md

This commit is contained in:
Joseph Henry 2015-10-27 17:41:51 -07:00
parent 18db95d0e9
commit 21889305e0

View File

@ -125,24 +125,28 @@ To Test:
### Extended Version Notes ### Extended Version Notes
20151027 Added Redis-server support 20151027 Added Redis-server support:
Notes:
- Added extra logic to detect socket re-issuing and consequent service-side double mapping. - Added extra logic to detect socket re-issuing and consequent service-side double mapping.
Redis appears to try to set its initial listen socket to IPV6 only, this currently fails. As Redis appears to try to set its initial listen socket to IPV6 only, this currently fails. As
a result, Redis will close the socket and re-open it. The server will now test for closures a result, Redis will close the socket and re-open it. The server will now test for closures
during mapping and will eliminate any mappings to broken pipes. during mapping and will eliminate any mappings to broken pipes.
20151021 Added Node.js support 20151021 Added Node.js support:
Notes:
- syscall(long number, ...) is now intercepted and re-directs the __NR_accept4 call to our intercepted accept4() function - syscall(long number, ...) is now intercepted and re-directs the __NR_accept4 call to our intercepted accept4() function
- accept() now returns -EAGAIN in the case that we cannot read a signal byte from the descriptor linked to the service. This - accept() now returns -EAGAIN in the case that we cannot read a signal byte from the descriptor linked to the service. This
is because the uv__server_io() function in libuv used by Node.js looks for this return value upon failure, without it we is because the uv__server_io() function in libuv used by Node.js looks for this return value upon failure, without it we
were observing an innfinite loop in the I/O polling code in libuv. were observing an innfinite loop in the I/O polling code in libuv.
- accept4() now correctly sets given flags for descriptor returned by accept() - accept4() now correctly sets given flags for descriptor returned by accept()
- setsockopt() was modified to return success on any call with the following conditions: - setsockopt() was modified to return success on any call with the following conditions:
level == IPPROTO_TCP || (level == SOL_SOCKET && option_name == SO_KEEPALIVE) level == IPPROTO_TCP || (level == SOL_SOCKET && option_name == SO_KEEPALIVE)
This might be unnecessary or might need a better workaround This might be unnecessary or might need a better workaround
- Careful attention should be given to how arguments are passed in the intercepted syscall() function, this differs for - Careful attention should be given to how arguments are passed in the intercepted syscall() function, this differs for
32/64-bit systems 32/64-bit systems