Summary:
This moves the load of favicons into the content process. We use the same logic
for finding favicons (based on waiting until none have shown up for a short
time) but then load the favicon and convert it to a data uri which we then
dispatch to the parent process. Along the way this fixes asssociating the load
with the tab for WebExtension and devtools, fixes CSP usage for the load, fixes
expiry detection of the favicon and stops us from loading the same resource
twice.
This change also merges the prefs browser.chrome.site_icons and
browser.chrome.favicons leaving just the former controlling favicon loading. It
adds the pref browser.chrome.guess_favicon to allow disabling guessing where
a favicon might be located for a site (at <hostname>/favicon.ico). This is
mainly to allow disabling this in tests where those additional yet automatic
requests are uninteresting for the test.
There are multiple clean-ups that can follow this but this is a first step along
that path.
MozReview-Commit-ID: E0Cs59UnxaF
Reviewers: mak
Tags: #secure-revision
Bug #: 1453751
Differential Revision: https://phabricator.services.mozilla.com/D1672
Differential Revision: https://phabricator.services.mozilla.com/D1673
Differential Revision: https://phabricator.services.mozilla.com/D1674
Differential Revision: https://phabricator.services.mozilla.com/D1850
Differential Revision: https://phabricator.services.mozilla.com/D1869
Summary:
This moves the load of favicons into the content process. We use the same logic
for finding favicons (based on waiting until none have shown up for a short
time) but then load the favicon and convert it to a data uri which we then
dispatch to the parent process. Along the way this fixes asssociating the load
with the tab for WebExtension and devtools, fixes CSP usage for the load, fixes
expiry detection of the favicon and stops us from loading the same resource
twice.
This change also merges the prefs browser.chrome.site_icons and
browser.chrome.favicons leaving just the former controlling favicon loading. It
adds the pref browser.chrome.guess_favicon to allow disabling guessing where
a favicon might be located for a site (at <hostname>/favicon.ico). This is
mainly to allow disabling this in tests where those additional yet automatic
requests are uninteresting for the test.
There are multiple clean-ups that can follow this but this is a first step along
that path.
MozReview-Commit-ID: E0Cs59UnxaF
Reviewers: mak
Tags: #secure-revision
Bug #: 1453751
Differential Revision: https://phabricator.services.mozilla.com/D1672
Differential Revision: https://phabricator.services.mozilla.com/D1673
Differential Revision: https://phabricator.services.mozilla.com/D1674
Differential Revision: https://phabricator.services.mozilla.com/D1850
The "pageshow" and "blur" event listeners in LoginManagerContent only matter
once the module has loaded and processed other events. Before that, they're
guaranteed to be no-ops.
This patch delays adding those listeners before LoginManagerContent is used
for a given frame script.
MozReview-Commit-ID: 1f5AOkRkAhp
The "pageshow" and "blur" event listeners in LoginManagerContent only matter
once the module has loaded and processed other events. Before that, they're
guaranteed to be no-ops.
This patch delays adding those listeners before LoginManagerContent is used
for a given frame script.
MozReview-Commit-ID: 1f5AOkRkAhp
This removes the bookmark description UI, but leaves the backend in-place for a release or two, so that users have time to save anything they want (e.g. via backup).
The backend will be removed in bug 1402890.
MozReview-Commit-ID: La4AYFar7eK
There are some places where we have a thing which may not even be a node, and
we end up hardcoding the value of DOCUMENT_NODE there, because
"foo.nodeType == foo.DOCUMENT_NODE" will test true if foo is not a node: both
sides will be undefined.
In addition to the code going to a .jsm, it became a singleton that implements nsIMessageListener, instead of being one instance per tab (which is usually done to give easier access to the page's content, but it's possible to get to it through each message, as message.target.content
This also removes any redundant Ci.nsISupports elements in the interface
lists.
This was done using the following script:
acecb401b7/processors/chromeutils-generateQI.jsm
MozReview-Commit-ID: AIx10P8GpZY
This patch moves all TLS error string handling to the frontend.
Dev-tools doesn't show the same error code as the page does anymore but only the error code as string.
All logging of these error messages has been removed.
Bug #: 1415279
Differential Revision: https://phabricator.services.mozilla.com/D607
This adds workarounds to ensure that messages passed from browser.js
to content.js in the context of certerror pages always contain a frameId
which can be used to identify the frame that is supposed to receive them.
This fix is really meant to be temporary until we come up with a good
replacement for chrome - content communication, which probably boils down
to finding a middle ground between nsAboutCapabilities, RemotePageManager and WebChannel.
I did not update communication for Captive Portal pages, since those require
one-way broadcasting from chrome to content, which is not supported in this model.
This is tracked in bug 1446319.
I did also not change the behavior of the "Go Back" button, which still navigates
away the top level page, because I consider changing that behavior out of scope
for this bug (and in my personal opinion we should not change the behavior).
MozReview-Commit-ID: GrM6PFys6Cu
This adds workarounds to ensure that messages passed from browser.js
to content.js in the context of certerror pages always contain a frameId
which can be used to identify the frame that is supposed to receive them.
This fix is really meant to be temporary until we come up with a good
replacement for chrome - content communication, which probably boils down
to finding a middle ground between nsAboutCapabilities, RemotePageManager and WebChannel.
I did not update communication for Captive Portal pages, since those require
one-way broadcasting from chrome to content, which is not supported in this model.
This is tracked in bug 1446319.
I did also not change the behavior of the "Go Back" button, which still navigates
away the top level page, because I consider changing that behavior out of scope
for this bug (and in my personal opinion we should not change the behavior).
MozReview-Commit-ID: GrM6PFys6Cu
The context-menu change is technically not idempotent, since something like:
background-image: url(foo), linear-gradient(..);
Would throw before. But that didn't seem like a great deal to me.
MozReview-Commit-ID: 70pD1EyXDB
This patch was autogenerated by my decomponents.py
It covers almost every file with the extension js, jsm, html, py,
xhtml, or xul.
It removes blank lines after removed lines, when the removed lines are
preceded by either blank lines or the start of a new block. The "start
of a new block" is defined fairly hackily: either the line starts with
//, ends with */, ends with {, <
