Fix memory leak when AICH hashing already known files
(upstream git commit 9e62350fae9f24de64987a0cb002fdc15b5fa9af)
Like FAT, NTFS doesn't like special characters either
(upstream git commit 41113ecf15019301afea6cb35d9c35a7b8a0bdd1)
Until now, a little-endian sender worked, regardless of the endianness
of the receiver, but a big-endian sender sent all integer data reversed.
(upstream git commit 07a5a373ff356166207438b5dffdf4eee30825d1)
Fix: first check our bool variable, then make some calculations if
it's false. Not the other way around
(upstream git commit 41090dfe3756396a8b4496f732ab8493aa51fe6c)
Fix: disabling protocol obfuscation broke Kad and triggered assertions
(upstream git commit ec2e66216738f92724a37fa030a79734e0e8b1ba)
Fix: disable the partfile importer's 'Add' button on the remote gui
unless using a localhost connection
(upstream git commit 004f3929d91e4b81f47235060d86686fff13e2d3)
Fix: "Prompt on exit" preference
(upstream git commit aea8d002b6446dcaf38db4d4ab222fcebce60948)
Fix: require restart when protocol obfuscation setting gets changed
(upstream git commit ec2e66216738f92724a37fa030a79734e0e8b1ba)
And fix libX11 '-lX11' linkage.
Activate a lot of useful stuff, like the amule daemon and the amulecmd
line tool, which makes it possible to run amule mostly unattended.
(you might need to run it with the gui to configure it at first).
Even the experimental amulegui appears to be mostly functional.
the webgui is included for reference, it works on i386, something fishy
on sparc64...
todo: figure out cross-platform issues. The people who write amule never
thought that the client and the server could run on distinct architectures,
and so they transfer binary data without even putting it in network format,
leading to interesting transfer times and sizes...
