With the old process selector service implementation, non-cached loads
would skip the call into the process selector.
This implementation piggybacks atop the existing 'redirectTo' mechanism.
This is unfortunately not the perfect system for catching these loads,
as it doesn't provide an opportunity for performing a final process
switch before redirecting to a non-http channel. In addition, it doesn't
provide indication that a redirect is going to happen, causing
potentially unnecessary process switches.
Not all places where 'redirectTo' is supported use this mechanism. This
process switching mechanism is only checked in situations after
http-on-examine-response.
Potential future changes include:
1. Moving these checks closer to the real 'OnStartRequest' call (e.g.
in ContinueProcessNormal + ContinueOnStartRequest3). This would mean
that loads other than the final load will not cause process swaps.
2. Adding a callback before a redirect is performed, passing in the new
channel, and allowing modifications to be made. This would allow
performing a process swap before redirecting to a non-http(s)
channel.
Depends on D15609
Differential Revision: https://phabricator.services.mozilla.com/D15610
For navigation: text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8
For images: image/png,image/svg+xml,image/*;q=0.8,*/*;q=0.5
for style: text/css,*/*;q=0.1
Anything else: */*
Summary: Really sorry for the size of the patch. It's mostly automatic
s/nsIDocument/Document/ but I had to fix up in a bunch of places manually to
add the right namespacing and such.
Overall it's not a very interesting patch I think.
nsDocument.cpp turns into Document.cpp, nsIDocument.h into Document.h and
nsIDocumentInlines.h into DocumentInlines.h.
I also changed a bunch of nsCOMPtr usage to RefPtr, but not all of it.
While fixing up some of the bits I also removed some unneeded OwnerDoc() null
checks and such, but I didn't do anything riskier than that.
WPT parses HTTP log and create a HAR file. To reduce the overhead of logging, this patch moves some logs that are used by WPT parser to level 1.
Differential Revision: https://phabricator.services.mozilla.com/D8986
Necko does not allow for a CONNECT only request to happen. This adds a flag
to signal an http channel should only CONNECT if proxied. This flag can only be
set if an HTTPUpgrade handler has been assigned. As proposed by rfc7639, an
alpn header will be included in the CONNECT request. Its value is determined by
the upgrade protocol passed to HTTPUpgrade.
The flag is added as part of nsIHttpChannelInternal because it is dependent on
HTTPUpgrade. It exists as a capability flag since the channel's transaction
needs to know to complete after a successful CONNECT. Also the transaction's
connection needs to know to stop writing transaction data after it CONNECTs or
if there's no proxy, and to not tell the transaction to reset.
If an nsHttpChannel has this flag set then the upgrade handler will receive the
socket after the CONNECT succeeds without doing tls if https.
In order to support xpcshell-test for this change, nsHttpConnectionMgr does a
little check to see if HTTPUpgrade callback is in JavaScript. If it is then
the callback is invoked on the main thread.
In the current code there are 3 main issues:
1. nsFileStream is not really thread-safe. There is nothing to protect the
internal members and we see crashes.
2. nsPipeInputStream doesn't implement ::Seek() method and that caused issues
in devtools when a nsHttpChannel sends POST data using a pipe. In order to fix
this, bug 1494176 added a check in nsHttpChannel: if the stream doesn't
implement ::Seek(), let's clone it. This was an hack around nsPipeInputStream,
and it's bad.
3. When nsHttpChannel sends POST data using a file stream, nsFileStream does
I/O on main-thread because of the issue 2. Plus, ::Seek() is called on the
main-thread causing issue 1.
Note that nsPipeInputStream implements only ::Tell(), of the nsISeekableStream
methods. It doesn't implement ::Seek() and it doesn't implement ::SetEOF().
With this patch I want to fix point 2 and point 3 (and consequentially issue 1
- but we need a separate fix for it - follow up). The patch does:
1. it splits nsISeekableStream in 2 interfaces: nsITellableStream and
nsISeekableStream.
2. nsPipeInputStream implements only nsITellableStream. Doing this, we don't
need the ::Seek() check for point 2 in nsHttpChannel: a simple QI check is
enough.
3. Because we don't call ::Seek() in nsHttpChannel, nsFileStream doesn't do I/O
on the main-thread, and we don't crash doing so.
This patch changes the way we set and handle the preferred alternate data type.
It is no longer just one choice, but a set of preferences, each conditional
on the contentType of the resource.
For example:
var cc = chan.QueryInterface(Ci.nsICacheInfoChannel);
cc.preferAlternativeDataType("js-bytecode", "text/javascript");
cc.preferAlternativeDataType("ammended-text", "text/plain");
cc.preferAlternativeDataType("something-else", "");
When loaded from the cache, the available alt-data type will be checked against
"js-bytecode" if the contentType is "text/javascript", "ammended-text" if the contentType is "text/plain" or "something-else" for all contentTypes.
Note that the alt-data type could be "something-else" even if the contentType is "text/javascript".
The preferences are saved as an nsTArray<mozilla::Tuple<nsCString, nsCString>>.
Differential Revision: https://phabricator.services.mozilla.com/D8071
This patch changes the way we set and handle the preferred alternate data type.
It is no longer just one choice, but a set of preferences, each conditional
on the contentType of the resource.
For example:
var cc = chan.QueryInterface(Ci.nsICacheInfoChannel);
cc.preferAlternativeDataType("js-bytecode", "text/javascript");
cc.preferAlternativeDataType("ammended-text", "text/plain");
cc.preferAlternativeDataType("something-else", "");
When loaded from the cache, the available alt-data type will be checked against
"js-bytecode" if the contentType is "text/javascript", "ammended-text" if the contentType is "text/plain" or "something-else" for all contentTypes.
Note that the alt-data type could be "something-else" even if the contentType is "text/javascript".
The preferences are saved as an nsTArray<mozilla::Tuple<nsCString, nsCString>>.
Differential Revision: https://phabricator.services.mozilla.com/D8071
The mIsTrackingResource flag on nsIHttpChannel was split into two separate
flags depending on whether or not the resource is third-party. The correct
flag will be set by the channel classifier. Similarly, a new function was
introduced, GetIsThirdPartyTrackingResource(), for those consumers (like TP)
who only care about third-party trackers.
The existing function, GetIsTracking(), will continue to look at both
first-party and third-party trackers (the behavior since first party
tracking was added to annotations in bug 1476324).
The OverrideTrackingResource() function now allows nsHTMLDocument to
override both mIsFirstPartyTrackingResource and
mIsThirdPartyTrackingResource, but since this function is a little dangerous
and only has a single user, I added an assert to make future callers think
twice about using it to opt out of tracking annotations.
Currently, only the default storage restrictions need to look at first-party
trackers so every other consumer has been moved to
mIsThirdPartyTrackingResource or GetIsThirdPartyTrackingResource().
This effectively reverts the third-party checks added in bug 1476715 and
replaces them with the more complicated check that was added in bug 1108017.
It follows the approach that Ehsan initially suggested in bug 1476715. It
also reverts the changes in the expected values of the tracking annotation
test since these were, in hindsight, a warning about this regression.
Depends on D3722
Differential Revision: https://phabricator.services.mozilla.com/D3723
Correctness improvements:
* UTF errors are handled safely per spec instead of dangerously truncating
strings.
* There are fewer converter implementations.
Performance improvements:
* The old code did exact buffer length math, which meant doing UTF math twice
on each input string (once for length calculation and another time for
conversion). Exact length math is more complicated when handling errors
properly, which the old code didn't do. The new code does UTF math on the
string content only once (when converting) but risks allocating more than
once. There are heuristics in place to lower the probability of
reallocation in cases where the double math avoidance isn't enough of a
saving to absorb an allocation and memcpy.
* Previously, in UTF-16 <-> UTF-8 conversions, an ASCII prefix was optimized
but a single non-ASCII code point pessimized the rest of the string. The
new code tries to get back on the fast ASCII path.
* UTF-16 to Latin1 conversion guarantees less about handling of out-of-range
input to eliminate an operation from the inner loop on x86/x86_64.
* When assigning to a pre-existing string, the new code tries to reuse the
old buffer instead of first releasing the old buffer and then allocating a
new one.
* When reallocating from the new code, the memcpy covers only the data that
is part of the logical length of the old string instead of memcpying the
whole capacity. (For old callers old excess memcpy behavior is preserved
due to bogus callers. See bug 1472113.)
* UTF-8 strings in XPConnect that are in the Latin1 range are passed to
SpiderMonkey as Latin1.
New features:
* Conversion between UTF-8 and Latin1 is added in order to enable faster
future interop between Rust code (or otherwise UTF-8-using code) and text
node and SpiderMonkey code that uses Latin1.
MozReview-Commit-ID: JaJuExfILM9
