| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
| |
This introduces a GeditOverlayChild subclass, listening to parent
overlay mouse events to define the escaping policy.
The distance from which the widget "escapes" the mouse pointer
can be set at construction time.
https://bugzilla.gnome.org/show_bug.cgi?id=643909
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
When a new window is opened and given a specific size and position by
javascript, the request should only be respected if the new window is
being opened stand alone, and considered a popup. This fixes bad
behaviour observed when the new window became a tab on an existing
window and messed with its size.
Bug #612155
|
| |
|
|
|
| |
It's where it belongs, and it will make things easier for the
following patches in this area.
|
| |
|
|
|
|
|
| |
WebKit 1.4.0 checks if symbols are being mixed before loading a
plugin. That is more robust, so we should rely on it.
Bug #647096
|
| |
|
|
|
| |
This still has some issues, but until we figure them out I guess it's
much better to turn the preference on than off.
|
| |
|
|
| |
This requires WebKitGTK+ r81486, soon to be 1.3.13.
|
| |
|
|
|
|
|
| |
To match the behavior of other browsers, the default web content font
is the serif font, instead of the sans-serif font.
Bug #644523.
|
| |
|
|
|
|
|
|
| |
Instead of using the Gnome interface font as the default serif font, use the
Gnome document font. The document font has an increased chance of being a serif
font over the interface font.
Bug #644521
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
EphyDownload is a wrapper object around WebKitDownload that handles common
behavior in downloads: auto-destination, default action for the MIME type.
It can be used to wrap a WebKitDownload coming from a WebKitView or to download
a url: ephy_download_new_for_uri and ephy_download_new_for_download are
provided.
Its lifetime is not automagic like EphyEmbedPersist, so you have to unref it
when you no longer need it.
This new object replaces EphyEmbedPersist and enables us to use a single
codepath for downloads in all Epiphany.
Bug #618443
|
| |
|
|
| |
Otherwise it blends a bit too effectively into clear backgrounds.
|
| |
|
|
|
|
| |
Since that's what we add to the overlay, its visible property will be
in sync with the child container. Gets rid of bothering 1px visible
leftover when statusbar is hidden.
|
| |
|
|
|
|
| |
Fixes a crash when trying to open the WebInspector
https://bugzilla.gnome.org/show_bug.cgi?id=642306
|
| |
|
|
| |
And adapt to the new API.
|
| |
|
|
| |
We now reach to the EphyEmbed, so this is dangerous.
|
| | |
|
| |
|
|
|
| |
Seems GeditOverlay works better if we pack directly the scrollable
widget on it, then pack the overlay in a scrolled window. Do that.
|
| |
|
|
|
|
|
|
|
|
| |
Works correctly in framed pages, and fixes a few bugs along the way
(not to mention it should be more efficient since it does not redraw
needlessly).
The gedit code has been modified to get rid of the animation stuff we
don't really need atm; we have coordinated with the gedit developers
and hopefully both versions will be in sync again really soon.
|
| |
|
|
|
|
| |
and update configure.ac to require the corresponding libsoup
https://bugzilla.gnome.org/show_bug.cgi?id=636861
|
| |
|
|
|
|
|
|
|
| |
There might be more than one plugin instance for flash. Iterate the whole list
to be sure we have disabled all the possible instances.
Bug #637959
Signed-off-by: Diego Escalante Urrelo <descalante@igalia.com>
|
| |
|
|
|
|
|
| |
Until 2.99.0 GtkStyleContext didn't have a "normal" flag, the closest thing was
GTK_STATE_FLAG_ACTIVE, but it is not necessarily correct.
Bug #636501
|
| |
|
|
|
|
|
| |
Updates all our uses of GtkStyle stuff to the newer GtkStyleContext API.
ephy-web-view porting done by Matthias Clasen <mclasen@redhat.com>
Bug #636501
|
| | |
|
| | |
|
| |
|
|
|
|
| |
The Adobe flash plugin uses GTK+2.x, so it cannot be loaded in our
(GTK+3.x) process without causing a crash. Disable it by default until
we implement a solution.
|
| | |
|
| |
|
|
|
|
| |
Show nice error pages instead of WebKitGTK+'s defaults.
Bug #592667
|
| |
|
|
|
|
| |
Non prefixed names trigger lots of warnings, avoid them.
Bug #636790
|
| |
|
|
|
|
| |
Prefix them with EPHY_PERMISSION_TYPE instead of "EPT_".
Bug #636790
|
| |
|
|
| |
Bug #636790
|
| |
|
|
|
|
|
|
| |
Allow users to set custom sans, serif and monospace fonts if they don't want
the defaults, which are the desktop wide settings in the
org.gnome.desktop.interface schema.
Bug #636761
|
| |
|
|
|
|
| |
minimum-logical-font-size doesn't works as expected by the UI.
Bug #636761
|
| |
|
|
| |
We can use the arguments of the function safely.
|
| |
|
|
| |
Settings kept in sync with g_signal_connect were missing an initialization.
|
| |
|
|
|
|
|
|
|
|
|
| |
The migration code is now in the ephy-profile-migrator binary, this means:
- epiphany is not linking to NSS anymore
- lib/ephy-profile-migration.c was split into:
+ lib/ephy-profile-migrator.c: the new ephy-profile-migrator binary
+ lib/ephy-profile-utils.c: convenience _ephy_profile_*_form_auth_data functions.
- testing for migration is done without running ephy-profile-migrator
Bug #636685
|
| |
|
|
|
|
|
|
| |
The inspector embedding breaks some assumptions about the internal embed
widgetry, mainly breaking EPHY_GET_EMBED_FROM_EPHY_WEB_VIEW macro.
The embed now has the main EphyWebView in a GtkPaned permanently.
Bug #628459
|
| |
|
|
|
|
| |
and control they are caused by the main frame.
Bug #623012
|
| |
|
|
|
|
| |
Some DOM objects are not garbage collected by WebKit (since they are
not in the DOM tree), manually unref them when they are not needed
anymore.
|
| |
|
|
| |
Bug #635066
|
| | |
|
| |
|
|
|
|
|
| |
As a FIXME we could support both (at the same time) somewhat easily,
we just need to add the needed APIs in WebKitGTK+.
Bug #615362
|
| |
|
|
| |
They are not even emitted anymore, and are unused.
|
| |
|
|
|
| |
We are going to revamp the downloads UI to fit better in the Shell
design, but for now we can just get rid of this.
|
| |
|
|
|
| |
I don't think it makes any sense to make this optional, much less to
have a UI preference for it.
|
| |
|
|
| |
Instead rely on notify::{h,v}adjustment
|
| | |
|
| |
|
|
|
|
| |
Adds our own schemas, a migration file and removes old gconf API and files.
Bug #624485
|
| |
|
|
|
| |
Mostly moving from ::expose-event to ::draw, and adapting to the
changes in the paint functions.
|
| | |
|
| |
|
|
| |
Bug #618627
|
| |
|
|
|
|
|
| |
Refuse to store in the cache any invalid icon type, this prevents us from
storing 404 html pages as favicons.
Bug #618627
|
| |
|
|
|
|
|
| |
When calling ephy_favicon_cache_get we might hit an invalid file type that we
didn't catch previously, delete it.
Bug #618627
|
| |
|
|
| |
Bug #618627
|
| |
|
|
|
|
| |
We just use it to update the buttons 100ms later, so having one is
enough. If we overwrite it we might crash down the road when we run
the timeout on a disposed DV.
|
| | |
|
| |
|
|
|
|
|
|
|
| |
Using the infrastructure for info bars, present policy requests to the
user so they can allow/deny them based on the URI of the frame that is
requesting. Moving forward, it would be a good idea to offer to save
the decision, and store it in the history entry for that site.
Bug #626687
|
| |
|
|
|
|
|
| |
Its current size is not very helpful, and causes a small part of the
inspector to be cut off horizontally.
Bug #626625
|
| |
|
|
| |
Bug #626493
|
| |
|
|
| |
Bug #626489
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This makes language autodetection, as implemented by some websites, to
work.
Most of the code stolen from libsoup.
Original patch from Mario Sánchez Prada.
Updated to work with non-English locales.
Approved by Gustavo Noronha.
Bug 602547
|
| |
|
|
| |
Bug #626454
|
| |
|
|
| |
Bug #625640
|
| |
|
|
| |
Bug #625922
|
| |
|
|
| |
Bug #625808
|
| |
|
|
| |
Bug #625404
|
| |
|
|
| |
Bug #601312
|
| |
|
|
| |
Bug #621050
|
| |
|
|
|
|
| |
Avoid duplication of code.
Bug #621050
|
| |
|
|
| |
Bug #609817
|
| |
|
|
|
|
|
|
|
| |
We were replacing the address with the old one even in the case
the user cancelled the load because he added a new web page before
the previous one finished the loading process. This patch adds a check
to control if we can expire the address.
Bug #620129
|
| |
|
|
|
| |
Would duplicate the functionality of the WebKit DOM signals, if it
worked at all...
|
| | |
|
| | |
|
| |
|
|
|
| |
There's no need anymore for the EphyWindow to do this, we can get rid
of one extra roundtrip.
|
| |
|
|
|
| |
Define them in the EphyWebView header, since we'll need to create the
context ids from multiple files.
|
| |
|
|
|
|
|
|
|
|
| |
Get rid of our statusbar for good and switch to something like what
Chromium uses, since it takes less vertical space.
The only regression is that we lose the resize grip, but that should
be re-added to GtkWindow soon.
Bug #609713
|
| |
|
|
|
| |
Trying to load file:, for instance, will crash Epiphany. This commit
fixes this. See http://bugs.debian.org/554595.
|
| |
|
|
| |
We removed the sidebar extension for 2.30.
|
| |
|
|
| |
file name
|
| |
|
|
|
|
| |
In this case the user cancelled the load, we set the address of the
current page loaded in the webview again in order to avoid showing
an URL that was not committed.
|
| |
|
|
|
|
|
| |
Modified the state of the webivew in case the load failed to avoid
keeping the old title or icon in that situation.
Bug #593743
|
| |
|
|
|
|
|
|
| |
Avoid setting the provisional uri as the title if there is already
a typed address, and set the loading_title before using it in the
status message.
Bug #593743
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
Create a method to make the EphyWebView load the homepage set by the
user. This is in preparation for creating a signal for this action,
which other code in Epiphany will need.
|
| |
|
|
|
|
| |
Make url_is_empty a helper function in embed/, since we'll be moving
code from EphyShell into EphyEmbed that uses it and we don't want to
duplicate it.
|
| | |
|
| | |
|
| |
|
|
|
|
| |
Does not really make sense for the EphyEmbed container to connect to
its child signal and then update its title, let the child itself
handle it.
|
| |
|
|
|
|
|
|
| |
This patch uses the ephy-embed.c callback code and refactors it in
just one method in the ephy-web-view that handles all the status
changes for this object.
Bug #593743
|
| |
|
|
|
|
|
| |
We can use the webkit load status (WebKitLoadStatus) and avoid
defining our own enum to check the net states.
Bug #593743
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
Added new function in EphyWebView to clear the history from
WebKitWebView, and connect to the 'cleared' signal in EphyEmbed to call
to such a function when needed.
Bug #539716
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
|
|
|
| |
Make sure the EphyEmbedPersist object is created specifying the
EphyEmbed object, and create the network request inside of it
using the URL from the associated web view as 'Referer'
Bug #136292
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
| |
Bug #582629
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
|
| |
We only accept lowercase schemes right now, for no good reason.
Bug #605089
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
|
|
| |
Use the already existing functions we have for print preview also for
printing; fixes a bunch of usability issues.
Bug #609756
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
| |
Changed the "Open" button to "Download" when the mime type of the file
is not considered safe for opening.
Bug #361559
|
| |
|
|
|
|
|
|
|
|
|
|
| |
When using notifications for downloads we usually hit the ugly case where the
notification pops but there is no GtkStatusIcon for it so it will pop in the
default location (usually bottom right). This is inconsistent with the behavior
when the GtkStatusIcon is present, which is to show it attached to it.
To fix this we hold a reference to the DownloaderView when showing the
notification and release it when such notification has been closed.
Bug #611779
|
| |
|
|
|
|
|
|
|
|
|
| |
The DV follows the same goal as before (being alive while there are downloads).
Now every download holds a reference to the DV and releases it when the
download is removed (finished, cancelled, etc).
The DV clears its own reference once and only once when a download is removed
and there are no more downloads in queue.
Bug #611779
|
| |
|
|
|
|
|
|
|
|
| |
When adding a download we should call update_download_row() *after* connecting
the signal handlers or they will never be disconnected.
Also, in update_download_row() we should consider that the case where it is
called with a download that has already been removed as a non fatal case (i.e.
just return;). This is because in downloader_view_add_download() we can have
the signal handlers remove the download from the view before the manual call
(it was happening vice versa before this).
|
| |
|
|
|
|
|
|
| |
When there are no more downloads in queue, hide the view. Also if a download is
added but finishes before showing the window, don't show the window nor the
notification.
Bug #611779
|
| |
|
|
|
|
| |
A minor simplification.
Bug #611779
|
| |
|
|
|
|
| |
The previous code wasn't disconnecting signals at all.
Bug #611779
|
| |
|
|
|
|
|
| |
When there are no downloads left, the Pause/Cancel buttons should not be
sensitive.
Bug #611779
|
| |
|
|
| |
Bug #611779
|
| |
|
|
|
|
|
| |
This flag determines if a download should not appear in the DownloaderView
window. The default is to appear, like in gecko times.
Bug #611635
|
| |
|
|
|
|
|
| |
Otherwise we can hit a case where the DownloaderView no longer exists (and
hence the notification widget) because the removed download was the last one.
Bug #611635
|
| |
|
|
|
|
| |
Otherwise we will hit an assert.
Bug #611635
|
| |
|
|
|
|
|
|
|
|
| |
Some sites don't use a sane type attribute in their input tags. So for example
you get input tags without a type, or with multiple types. Offenders are
bugzillas and ohloh.net.
Luckily, WebKit inferres correctly what the input should be so instead of
relying on what the HTML says we rely on what WebKit thinks.
Bug #608740
|
| |
|
|
|
|
|
| |
This retrieves a property from a JS object, just like "element.property" in
plain javascript.
Bug #608740
|
| |
|
|
|
|
|
|
|
|
|
| |
If we handle shift+click in EphyWebView default handler then we will be
preventing WebKit from getting *any* shift+click event, this breaks shift+click
for extending selection. Removing that however makes shift+click browse to
links just as click would do. The solution is to prevent
policy_decision_required_cb to browse to the clicked link when it includes the
shift modifier.
Bug #610844
|
| | |
|
| |
|
|
|
|
|
|
| |
This uses a new feature in libsoup that reports through a
SoupMessageFlag whether the message is talking to a server that has a
trusted server.
Bug #600663
|
| |
|
|
|
| |
There's really no way to completely disable Java support alone, and I
think enabling/disabling plugins in general is much more useful.
|
| |
|
|
|
|
|
| |
This fixes a crash reported ad-hoc in
https://bugzilla.gnome.org/show_bug.cgi?id=609753. This is needed
because we may bail out early, and try to free these strings without
having properly initialized them.
|
| |
|
|
|
|
|
|
| |
It doesn't do what we expect it to do to keep the download object alive, a
simple g_signal_connect does the right thing instead.
This fixes the downloads not moving to final destination problem.
Bug #609263
|
| |
|
|
|
|
|
|
| |
In ephy_web_view_button_press_event we are being too liberal with "&" when
checking for modifiers, we want == so we don't prevent handling elsewhere.
Also explain the function a bit more.
Bug #604950
|
| |
|
|
| |
Too wide comments, some indentation.
|
| |
|
|
| |
Bug #604991
|
| |
|
|
|
|
|
|
|
| |
JSStringGetLength returns the number of UTF8 chars in the string, using this
value results in trimmed strings when they contain UTF8 chars.
The correct function is JSStringGetMaximumUTF8CStringSize that reports the
number of bytes the string requires to be allocated, including null-byte.
Bug #609936
|
| | |
|
| |
|
|
|
|
| |
A previous patch to fix it from crashing on NULL strings changed how
the strings are managed in general, so we were using variables
containing garbage in some cases.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Created function ephy_web_view_show_print_preview, which replaces the
old implementation of print preview, which was not working now.
Preview is displayed in an external viewer, so print preview mode does
no longer exist.
All functions of the old implementation of print preview have been
removed, PPViewToolbar was removed also. Also, as EphyWebView has no
more a print preview mode, all functions which checked if a view was
in print preview mode were modified.
Bug #609021
|
| |
|
|
|
| |
There's a couple of issues with it we need to figure out, so disable
it for the release.
|
| |
|
|
| |
Bug #608517
|
| | |
|
| |
|
|
| |
Bug #607484
|
| |
|
|
| |
Bug #608749
|
| |
|
|
|
|
|
|
| |
EphyEmbedPersist is the object in charge of most context menu options. It
handles "Save target as", "Open Image", "Save image as", among other actions.
This reimplements it using WebKitDownload.
Bug #600987
|
| |
|
|
| |
Since we are not marking windows as popups for now.
|
| |
|
|
|
| |
All critical bugs have been fixed, we already depend on a new enough
version of WebKit, and this feature needs testing.
|
| |
|
|
|
|
|
|
|
|
| |
Some resources might be of the style "page.php?orig=/place/other", this would
make Epiphany crash while saving them because of the invalid characters for the
filename.
To solve this we escape the name of the resources before using it as the
name of the destination file.
Bug #606876
|
| |
|
|
|
|
|
|
|
|
| |
In ephy_web_view_save_sub_resources the destination_uri var is passed down to
the GAsyncReadyCallbacks and is freed after all the operations are finished,
however it is named sub_destination_uri in this callbacks. To make it more
obvious that it is the same data we renamed all the ocurrences to
destination_uri.
Bug #606876
|
| |
|
|
| |
Bug #606876
|
| |
|
|
| |
Bug #605480
|
| | |
|
| |
|
|
|
|
|
| |
WebKit will stop emitting load signals when it loads the error page,
which makes life easier for us.
Bug #377012
|
| | |
|
| |
|
|
|
|
| |
Include a mention to the username being saved and the domain where the form is.
Bug #605022
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
If a download is finished while the GtkStatusIcon for DownloadView is not
embedded the finished notification for such download is missed and user is
left wondering what happened with his/her download.
This makes notifications show whether there is or not a GtkStatusIcon to
attach to.
Bug #604599
|
| |
|
|
|
|
|
| |
Always download to a temporary location, only move the file to the final
destination after it has been downloaded completely.
Bug #594192
|
| |
|
|
| |
Bug #594192
|
| |
|
|
| |
Bug #594192
|
| |
|
|
|
|
|
|
|
| |
Sometimes downloads can have a mime_description but a NULL helper_app
in that case action is never changed so DOWNLOAD_ACTION_DOWNLOAD remains
as action value. This is the same response value as Save as...
button, which is wrong for the Download button.
Bug #605091
|
| |
|
|
| |
Bug #605091
|
| |
|
|
|
|
| |
Honour our database of safe and unsafe mime types.
Bug #605026
|
| |
|
|
|
|
|
|
| |
We consider all windows opened by the web page without user
intervention as popups, and allow them to be closed in the same way if
the page so requests.
Bug #599009
|
| |
|
|
|
| |
This is so that changes to the item objects do not affect other
WebViews.
|
| |
|
|
| |
Makes WebKit apply a series of workarounds for known broken sites.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We default to the WebKitGTK+ user agent with Epiphany and the version
appended to it.
This can be overriden/modified in two ways:
- The gconf key /apps/epiphany/general/user_agent has the maximum
priority, and if it's set it will be always used.
- If the gconf key is not set, vendors can still customize the UA by
placing a branding.conf file in the share data directory. One
example would be:
[User Agent]
Vendor=SUSE
VendorSub=11.2
VendorComment=2.29.94-1
And the result would be:
Mozilla/5.0 (X11; U; Linux i686; fr-fr) AppleWebKit/531.2+ (KHTML, like Gecko)
Safari/531.2+ SUSE/11.2 (2.29.94-1) Epiphany/2.29.94
Bug #580903
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
|
| |
New handler added to watch for changes in the GConf key with the list
of languages for the "Accept-Language" HTTP header, and update the
SoupSession's "accept-language" property (since libsoup 2.29.4).
Bug #528893
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Fixes crasher in pages with no rel elements.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
Fixes crashers when we actually make them the default shortly
afterwards.
|
| | |
|
| |
|
|
| |
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
|
|
| |
This is a "draft" quality implementation. There are some FIXMEs, and a
bunch of caveats, like private profiles not being taken into
consideration, and the user not being asked whether they want to save
the passwords or not.
Bug #582267
|
| | |
|
| | |
|
| |
|
|
|
| |
Otherwise we need to query keyring all the time to figure out if we
have to prefill forms.
|
| |
|
|
|
|
|
| |
This allows applications to provide their own context menus, and
perform their own handling for right-click events.
Bug #603326
|
| |
|
|
|
|
| |
Fixes Amazon Add to Wishlist bookmark.
Bug #604489
|
| |
|
|
|
|
|
|
|
| |
This shouldn't happen in theory, but the double check fixes a crasher
opening URLs from gmail.
Bug #604375
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
| |
Also redundant since we have the same thing in WebKitWebView now.
|
| | |
|
| |
|
|
| |
We already have WebKitWebView::icon-loaded, so it's redundant now.
|
| |
|
|
|
|
|
|
| |
This restores basic favicon functionality. In the future we will want
to replace our favicon cache with the IconDatabase that is available
in WebKit.
Bug #601859
|
| |
|
|
|
|
|
|
| |
Defaults to Google, like it used to.
Bug #604087
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
| |
This allows us to perform downloads when the server tells us to do so,
even for types WebKit supports. This fixes downloading GMail attachments.
Bug #598605
|
| |
|
|
| |
Connecting to notify::document-type is enough.
|
| | |
|
| |
|
|
|
|
|
| |
WebKit still thinks of them as HTML documents, and otherwise things
like the text-encoding menu are disabled for them.
Bug #603927
|
| |
|
|
| |
Bug #503852
|
| |
|
|
| |
Bug #503852
|
| |
|
|
| |
Bug #503852
|
| |
|
|
| |
This is used by external download managers like gwget.
|
| |
|
|
| |
Bug #603324
|
| |
|
|
|
|
| |
Partial fix for bug #599757.
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
| |
Otherwise we will rewrite the right document type with the document
type from other frames in the page.
Bug #602400
|
| |
|
|
|
|
| |
Partial fix for bug #599757.
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
| |
Partial fix for bug #599757.
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
| |
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
| |
This brings back proper content type reporting, and the 'Open' button
in the download dialog.
Bug #598167
|
| |
|
|
|
|
|
|
|
| |
Stop having each EphyWindow ref the shell, and instead have the shell
track all the newly created EphyWindows. When the last one is gone,
quit the GTK+ mainloop. This is simpler and avoids potential reference
cycles (see bug #573551).
Bug #599348
|
| |
|
|
| |
Bug #528799
|
| |
|
|
|
|
| |
(Part of) Bug #584537
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
| |
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| |
|
|
|
|
|
| |
This change makes sure that all characters that are typed make it to
the search query. Thanks to Dan Winship for pointing out the solution.
Bug #596717
|
| |
|
|
|
|
|
| |
Also consider localhost being used along with a port, with a path, and
with a domain.
Bug #595690
|
| | |
|
| |
|
|
|
|
|
| |
When a page fails to load (for example, when the user cancels it) it
won't go through the FINISHED state, which is the only one we are
taking into account to decide if a page is still loading. Use the new
WEBKIT_LOAD_FAILED status (WebKitGTK+ r48719) too.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I've used all the GTK+ 2.17.11 api available, still missing:
GTK_WIDGET_REALIZED (GTK_WIDGET ())
GTK_WIDGET_MAPPED (GTK_WIDGET ())
GTK_ENTRY->im_context
GTK_STATUSBAR->frame
GTK_STATUSBAR->label
GTK_MESSAGE_DIALOG->label
https://bugzilla.gnome.org/show_bug.cgi?id=595791
Signed-off-by: Xan Lopez <xan@gnome.org>
|
| | |
|
| |
|
|
|
|
|
|
| |
Use '%s Files', instead of '%s/%s Files' as the translatable string to
avoid confusion, and reuse existing mozilla-era translations. Fix the
languages that had already translated the new strings using old po
files as reference. Also add a translator comment to the source file,
for future reference.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
There's a few items (like email link) and actions (like bookmark link)
missing or not working because of missing information in the
WebKitHitTestResult object, but most of the stuff is working.
For some reason the g-ir-scanner is not picking up the correct type
name for WebKitHitTestResult (it uses WebKitHitTestResult instead of
WebKit.HitTestResult), so the introspection support is broken unless
that error is fixed manually. Looking into that ...
Bug #562617
|
| |
|
|
| |
Bug #562611
|
| |
|
|
| |
By calling mozilla-plugin-config on startup.
|
| |
|
|
|
|
|
|
|
|
|
| |
Otherwise the typed address would get stuck if you tried to load a
page while another was already loading, ie:
- Load URL.
- While it's loading, Ctrl+L, type something else, Enter.
- The new page will load, but the typed address won't be replaced with
the full address as usual. This is particularly bad when you type a
few letters and complete the URL in the dropdown.
|
| |
|
|
| |
It was just another useless abstraction at this point.
|
| |
|
|
|
| |
Bug 585499 - [Webkit] Automatic downloading into ~/Downloads fails if file
already exists
|
| |
|
|
|
|
|
|
| |
Most downloads will only have proper suggested filenames after the
request has been sent, and a reply was received, so delay getting the
suggested filename to when that has happened.
Bug #577797
|
| | |
|
| |
|
|
| |
Bug #562762
|
| |
|
|
| |
It might go away for 2.30, so protect against that.
|
| |
|
|
|
|
| |
The whole thing just had one functionality at this point as far as I
can see: prevent the typed address from being wiped out when a page
is loading. Simplify the code to do just that.
|
| |
|
|
| |
That's what ephy/gecko used to do (and I don't like it).
|
| |
|
|
|
|
|
|
|
| |
preference
Almost the same than 'minimum-font-size', but broadly speaking WebKit
won't enforce it when it considers it could break the layout of a
site. Should avoid breakage in some sites when we have a minimum font
setting too big.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
Avoids having a blank URL entry until the server responds to our
request.
Bug #591294
|
| |
|
|
|
| |
This allows Epiphany itself, and extensions to treat keys that are not
handled by the WebView for in-page search purposes.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
When WebKitGTK+ fails to load a URI and is allowed to load the default
error page it will emit a COMMITTED load-status with the original URI
we failed to load. This was confusing epiphany, since in theory
COMMITTED should be only emitted with URIs known to be good. As a
workaround flag pages that fail to load in the callback for
'load-error' and avoid doing anything with them when they arrive in
COMMITTED.
Bug #593200
|
| |
|
|
| |
Just return the status we set when loading.
|
| |
|
|
| |
Failing to delete a favicon file should not cause the program to abort.
|
| |
|
|
|
|
|
| |
Use a regular expression and SoupURI to figure out if we should search
the string that was typed, or load it as an URI.
Bug #583795
|
| |
|
|
|
|
| |
Depends on WebKitGTK+ r47776 (will be in 1.1.14)
Bug #591072
|
| |
|
|
|
| |
Much more clear, and avoids confusions with the WebKitWebView function
with similar name.
|
| |
|
|
| |
Bug #584506
|
| |
|
|
|
| |
We depend on WebKitGTK+ trunk (r47129) and libsoup
master (f81520bfd3a97) now.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
EphyLink is called repeatedly when opening URLs, so normalizing the
address there all the time is pointless. Also, we don't go through
here in all cases, so we need to add the normalization in the
ephy_web_view_load_request function too; this fixes opening URLs like
"foo.com" from the command line, for example.
|
| | |
|
| |
|
|
|
|
|
|
| |
Apparently due to a GTK+/gnome-panel bug (bug #340110), the status icon space
is not freed in the notification area.
The best work-around is to hide the icon before unrefing it.
Bug #591189
|
| | |
|
| |
|
|
|
|
|
|
|
| |
Patch by Holger Freyther, Prit Laes, with fixes from Xan López.
Fetches the data from the GNOME keyring, which is where WebKit stores it
(when compiled with keyring support).
Bug #579217
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
This is a workaround for webkit bug
https://bugs.webkit.org/show_bug.cgi?id=26409.
FIRST_VISUALLY_NON_EMPTY_LAYOUT might be emitted after LOAD_FINISHED, so
we just ignore any status change once LOAD_FINISHED has been set
|
