New API method: WebGUI::Response::sendFile; it, as appropriate, calls
setRedirect or setStreamedFile, depending on enableStreamingUploads config var.
setStreamedFile now kicks off the XSendfile process.
File.pm now uses this instead of trying to set both a redirect and a stream.
IO::File::WithPath blows up if a file doesn't exist so this raises an exception
now.
The http now no longer insist that '0' is not a valid filename to stream.
site.psgi, depending on enableStreamingUploads, enables either the Static
or XSendfile middleware.
Move logic out of WebGUI::Session::HTTP and into WebGUI::Session::Response /
::Request; deprecate more functions; change references in core to use
$session->response instead; fix tests that broke because of the change but
not one that merely generate the deprecated warning because I want to know
that the proxying of depricated methods is working. These can be changed later.
a crutch for converting tests, though really, it's enough shorter that it probably should live on.
getPage should be deprecated and replaced with getPage2 which should then be renamed. Joy.
Anyway, WebGUI::Test::Mechanize had a bit of chicken-and-egg going on with not being able to
modify things in its session until after a request with a valid session cookie was made.
It's now more forgiving.
Reworked t/Asset/AssetExportHtml.t slightly to use getPage2 as it currently stands.
requests using Plack::Test and attempts to maintain getPage's API.
Change three tests in t/Asset/AssetExportHtml.t to use this instead.
Backstory:
Forthcoming WebGUI::Session::HTTP, ::Request, ::Responses mucking abouts broke
some tests that use WebGUI::Test::getPage; in the case of t/Asset/AssetExportHtml.t,
files were written correctly to disc but tests failed because it was comparing them
to undef, which it got back for the page, apparently because getPage bypasses the
logic that traps printing to a filehandle.
This is minimal in implementation right now in that there's no support for passing args, which would generally be objects
mixed with strings, to the middlewares being added, as would be necessary if the .psgi files were to be largely migrated
to the config files. I agree that it's important that upgrade scripts be able to add middleware (probably custom ones
that know how to get the session from %$env) but I'm not convinced of the utility of moving much logic out of
share/site.psgi.
that it understands HTML. This interacts badly with Plack::Test and doesn't have much of
a point in real life anyway, and I accidentally sent the test up that dends on this being
gone without sending this too. Argh. If I could send up a coherent commit just once...
making my Mozilla shoot to 1.5gigs and taking the server and browser forever to get the
error page up.
keep the HTML of the original page and tack it on to the end of the stack trace.
add WebGUI::Middleware::StackTrace to the default site.psgi in place of the standard StackTrace,
but leave a comment for the other one in case people prefer it.
