StylesheetContext::response_complete called multiple times w/ WebComponents.

[eddyb]

After #10247, Servo can load enough of https://elements.polymer-project.org/bower_components/paper-button/demo/index.html (or, a simpler example, https://elements.polymer-project.org/bower_components/iron-fit-behavior/demo/index.html) to result in the following panic:

(EDIT: N.B. This first panic is tracked by #12047, the second one is specific to this issue)

thread 'ScriptThread PipelineId { namespace_id: PipelineNamespaceId(0), index: PipelineIndex(0) }' panicked at
'assertion failed: count_cell.get() > 0', /home/eddy/Projects/servo/components/script/dom/document.rs:1144
stack backtrace:
   1:     0x55ed83bdc3a0 - sys::backtrace::tracing::imp::write::h495bc662e480d01f2cv
   2:     0x55ed83be12ef - panicking::default_handler::_$u7b$$u7b$closure$u7d$$u7d$::closure.44522
   3:     0x55ed83be0f68 - panicking::default_handler::h4e73712d1e927dfeH0z
   4:     0x55ed83bcacec - sys_common::unwind::begin_unwind_inner::h0e1be209f5e9e87dg2t
   5:     0x55ed8247c34f - sys_common::unwind::begin_unwind::h14217152112849155013
   6:     0x55ed825594b8 - dom::htmllinkelement::StylesheetContext.AsyncResponseListener::response_complete::h2eb8bd1165b1b970oU4
   7:     0x55ed82f2f6d3 - ResponseAction::process::heea94a2037600089Ypf
   8:     0x55ed82558187 - network_listener::ListenerRunnable<T>.Runnable::handler::h2666221587141280080
   9:     0x55ed82709894 - script_thread::ScriptThread::handle_msg_from_script::h0d5ea602ace550ebACt
  10:     0x55ed8274bc0b - script_thread::ScriptThread::handle_msgs::_$u7b$$u7b$closure$u7d$$u7d$::closure.188277
  11:     0x55ed8273a1b9 - script_thread::ScriptThread::handle_msgs::h18e431bcf8aa6677eot

By commenting out the body of Document::{in,de}crement_script_blocking_stylesheet_count, the underlying problem is exhibited:

thread 'ScriptThread PipelineId { namespace_id: PipelineNamespaceId(0), index: PipelineIndex(0) }' panicked at
'unknown completed load Stylesheet(Url { scheme: "https", scheme_data: Relative(RelativeSchemeData { username: "", password: None, host: Domain("fonts.googleapis.com"), port: None, default_port: Some(443), path: ["css"] }), query: Some("family=Roboto:400,300,300italic,400italic,500,500italic,700,700italic"), fragment: None })', ../src/libcore/option.rs:704
stack backtrace:
   1:     0x55b069c09460 - sys::backtrace::tracing::imp::write::h495bc662e480d01f2cv
   2:     0x55b069c0e3af - panicking::default_handler::_$u7b$$u7b$closure$u7d$$u7d$::closure.44522
   3:     0x55b069c0e028 - panicking::default_handler::h4e73712d1e927dfeH0z
   4:     0x55b069bf7dac - sys_common::unwind::begin_unwind_inner::h0e1be209f5e9e87dg2t
   5:     0x55b069bf8d58 - sys_common::unwind::begin_unwind_fmt::h45dc2fac9b5e0d23m1t
   6:     0x55b069c08761 - rust_begin_unwind
   7:     0x55b069c45dcf - panicking::panic_fmt::h40f5ec0cdc3fc429FRL
   8:     0x55b069c51004 - option::expect_failed::h56e104933d0889c7LpO
   9:     0x55b0684ab746 - document_loader::DocumentLoader::finish_load::hc65f67d3d717ad00Jdb
  10:     0x55b0684a998a - dom::document::Document::finish_load::hd5020bc39a09b698qEO
  11:     0x55b0685860c9 - dom::htmllinkelement::StylesheetContext.AsyncResponseListener::response_complete::h5538b238e4650b00xT4
  12:     0x55b068f5c793 - ResponseAction::process::heea94a2037600089Ypf
  13:     0x55b068585187 - network_listener::ListenerRunnable<T>.Runnable::handler::h14943577174353779088
  14:     0x55b068736854 - script_thread::ScriptThread::handle_msg_from_script::h93c0d0f15dd73f87JBt
  15:     0x55b068778bcb - script_thread::ScriptThread::handle_msgs::_$u7b$$u7b$closure$u7d$$u7d$::closure.188264
  16:     0x55b068767179 - script_thread::ScriptThread::handle_msgs::hd7d61674a04d5f4dnnt

[alex]

FWIW, https://www.ssllabs.com/ssltest/viewMyClient.html also triggers this.

[eddyb]

@alex Wow, that's good news, more reproductions! The only thing I notice on that page is:

<script async="" src="http://plaintext.ssllabs.com/plaintext/script.js?t=1464221349657&amp;_=1464221349650"></script>

So maybe async scripts are just broken? How can we confirm this?

[jdm]

With the changes from #12047 applied I can't reproduce the problem on any of the URLs listed here.

[jdm]

Sorry, I didn't notice that the original URL (rather than the simplified one) still shows the "unknown completed load" panic. Let's keep this issue focused on that panic and leave #12047 for the original assertion.

[nox]

I don't get a panic anymore.