| Commit message (Collapse) | Author | Age |
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
checking in windows. To do so, tested binaries must still be supplied
with valid commandline arguments, but on windows gnunet will utilize the
-d flag to run the programs initialization phase or privileged
operations only. In these modes, a program will not enter its mainloop
or communicate with the outside.
updated relevant function calls gnunet-wide to meet the extended
function parameters.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
This code is considered experimental
and has not yet undergone intensive testing. Use it at your own peril, but we would still be really glad for
feedback and bug reports! (thanks!)
* fixed a typo in the exit-makefile (buildflags)
* adjusted output of gnunet-helper-exit-windows to properly identify itself
* updated driver-paths for looking for tap32/tap64
|
| |
|
|
|
|
|
|
|
| |
there seems to be no point in adding support for partial read from the
TAP, as there is no delimiter and we get data on a per-frame level.
|
| |
|
|
|
|
|
|
| |
added check if we are running on a win32/wow64/win64 host system.
|
|
|
|
|
|
| |
removed a few obsolete variables
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
named pipes
moved declaration of hdr outside of the ifs
fixed an incorrect comparison introduced in last commit
|
|
|
|
|
|
|
|
|
| |
consistent
attept_write now properly resets the read facility to IOSTATE_RESUME, in
case read was stalled because write was busy
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
woops logics bug discovered: when read goes into waiting, the output
facility will
not transfer the input buffer and forget about it silently...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
newly set IP addresses now should expire upon reboot (buggy on some
windows, but at least it solves the IP-persistence issues in win7 and
above)
cloned read functions to properly apply or strip the gnunet message
header
removed the status BOOL from the overlapped struct and made it local
some fixes here and there
added a lot of comments
style adjustments
TODO:
* actually strip the gnunet message header or apply it again
* test with the main vpn binary
|
|
|
|
|
|
|
|
|
|
|
| |
* added functionality to remove ip addresses from our interfaces
* some formatting work
* added rety-loop for the interface-name resolver (retry up to 30
seconds)
* set new ipv4/v6 addresses to be forgotten on bootup (store=active)
* adjusted some debug output (added a few \n here and there)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Registry is extremely slow at refreshing. inserted a makeshift sleep(5)
to avoid retrieving outdated interface-information.
added proper return-codes to set_address4/6
fixed incorrect string-length handling for the HWID tags
updatedriverforplugandplaydevicesa now updates only the driver for our
current device, no longer the other siblings too (for >10 devices,
this could easily have taken > 60 seconds! ).
removed inet_pton, as we require API version XPSP2 anyway.
|
|
|
|
|
|
|
|
|
| |
mandatory when adding a new device node. unfortunately this is pretty
slow...
added new required include+library to link against: newdev.h/dll
|
| |
|
| |
|
|
|
|
|
|
|
| |
added teardown functionality.
now: testing...
|
|
|
|
|
|
|
|
|
| |
the same way as we can read from our TAP.
added high-level description of attempt_read
|
|
|
|
|
|
|
|
|
|
| |
handed down as pipes.
helper will reopen stdin and stdout in overlapped mode. ReOpenFile
requires WinXPSP2 or Win2003SP1.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
differently(in contrast to *nix). overlapping does not exist for console
handles, and different functions are required to read/write/poll
stdin/out.
- THus, we can now detect the type of handle we got as
stdin/stdout from our calling process.
- Did a bit of refactoring.
next: add logics to use the proper access mode depending on our handletype.
|
|
|
|
|
|
| |
more research on the conin/stdin+overlapping issue.
|
|
|
|
|
|
|
|
|
|
|
|
| |
* ready (to do work),
* queued (waiting for async-io to return),
* waiting (a read-facilities wait for its output partner to get process)
* failed (if a socket error occured)
added attempt_tap_read machine.
reworked run() for overlapped IO
|
| |
|
|
|
|
|
|
| |
* started work on our select-equivalent for windows.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Only version 9.8 and later support IPv6, 9.8 is broken with IPv4. Thus,
we are using the same minium version as openvpn 2.3: tap32-9.9
* created our TAP/TUN handle-object (equivalent to the *nix init_tun()
function), this still requires some more logics to be
complete.
* added logics to set the tun up (+a fixed sleep, to wait for it to
come up)
* we now also store the device's GUID, as we need it for creating the
handle.
* added the exports of tap32: src/include/tap_windows.h , for the sake
of upgradability. This file contains version-specific defines for the
driver.
* happy new year!
|
|
|
|
|
|
|
| |
according to recommendations found in the C/C++-specs(wchar !=
unicode support, but may break compiler compatibility).
|
| |
|
|
|
|
|
|
|
|
| |
netsh.
Added wrapper for launching programs in a windows shell.
Added define for _tpopen
|
|
|
|
|
|
|
|
| |
finding anything. Switched over to backup plan:
* lookup the devices's name directly in registry via the PNPInterfaceID. Registry fun, yey!
* added a few fixes here and there
|
|
|
|
|
|
| |
the registry. This information is needed for netsh later on.
|
|
|
|
|
|
|
|
|
|
|
| |
virtual interface again. Needed, because netsh requires us to resolve
the devices name as string, which can not be derived off the
deviceinfoset directly.
we are now using the handed over devicename + our PID
the result looks something like this: gnunet-vpn13381
|
|
|
|
|
|
| |
added some fixes here and there. added more comments to the code.
|
| |
|
|
|
|
|
|
| |
the win32 vpn-helper.
|
|
|
|
|
|
|
|
| |
win32 based on the info that can be found in the MSDN.
Current revision: http://msdn.microsoft.com/en-us/library/windows/hardware/ff549791%28v=vs.85%29.aspx
|
|
- modified vpn-makefile definitions to process gnunet-helper-vpn-windows
instead of the regular one when using mingw as toolchain
|