This revamps the way proxy handling is done to allow us to just proceed as
normal, and then if we get HTTP errors that indicate that we need a proxy or
need proxy credentials, we ask for them and then retry everything.
We still try to auto-detect the need for a proxy on our very first invocation
because on the first invocation, a failure to fetch a URL may well indicate
that a proxy is needed, but that same assumption does not hold on subsequent
invocations. In those later cases, it's probably just a transient network
failure and it would be weird and annoying to pop up the "please provide proxy
config" dialog in those cases.
Thanks pb00068 for getting this ball rolling, even though once again I have
nearly entirely rewritten the PR. I need to kick that habit. :)
What a horrible wart lingering from the early days of the Internet. Note to
future designers: please do not express configuration using a Turing complete
language.
Thanks jenriq for getting the ball rolling on this.
Closes#196.
This makes it *way* easier to view and edit the translation files and avoids
the danger of accidentally slipping UTF-8 encoded characters into existing
files during editing.
- remove hardcoded 60 seconds timeout during verify, as with pack200
this may fail on slow systems
- introduce new config "timeout_verify" with 60 seconds default
- handle timeout with aborting the launch
This also makes the use of a Proxy instance explicit in the Getdown code,
rather than trying to thread things through the proxy-related system
properties.
It introduces a ProxyAuth SPI which allows an app installation to save proxy
credentials in whatever way it deems fit. Otherwise the credentials are unsaved
and the user effectively has to enter them every time Getdown attempts to
access the internet.
This is obviously a terrible design, but there's no good (cross-platform) way
to securely store a username and password. So anyone using this feature is
likely going to need to provide an implementation of ProxyAuth that does
whatever auth storage they deem appropriate.
Adapted from PR from @ThomasG-AI, thanks!
Typical use case is a security patch, were a new java patch
(java_windows.jar) is put on server side and java_min_version updated
accordingly in getdown.txt
Currently the new virtual machine will be downloaded only after
a second getdown launch. This pull request fixes this.
Michael Bayne