| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Change-Id: I611f0f7e6eeea0247956e0f1a9f2ec3ec51c9eef
|
|
|
|
|
|
|
|
|
| |
New assets, new red color for drafts.
Super-collapsed becomes red when there's a draft.
Conversation footer is hidden when last message is a draft.
Change-Id: I89a58971a19475be79f56cae53934f3c7b2ab894
Missing: properly re-sizing the footer.
|
|
|
|
| |
Change-Id: I9eda1060f4e49a0448a321f603aec0a254daf86a
|
|
|
|
|
|
| |
b/15595931
Change-Id: I51b5ce1d9387d6354d421b0e399a2003c6e84c54
|
|
|
|
|
|
|
|
| |
Fixes b/14812974. Instead of converting (now) floats
to strings for passing via javascript, just pass them directly
as ints. Then we don't have to do any parsing.
Change-Id: Iaab27e46b3c6875fee203de389119b835fd27261
|
|
|
|
|
|
|
|
|
| |
Provides a space above the attachment bars to show
some additional content. Default behavior does nothing.
Hooks have been added to allow resizing of the message
footer.
Change-Id: Ib950091d176311bdb06c7c7a466223beb997a7b5
|
|
|
|
|
|
|
|
|
| |
There's a bug in chromium webview that causes the scroll
position to be improperly reported the first time
onScrollChanged is called. This breaks rotation
on K+ so we delay it until the content is ready on K.
Change-Id: I57729ad5279a999ab523919c6059897813b1370e
|
|
|
|
|
|
|
|
|
| |
Fixes b/13433568. dir="auto" uses the content
of its text to set the direction. This makes the
text show up on either the left or right aligned
depending upon locale.
Change-Id: Ifff7b6b7b6ffddbd386c14545bc094bbb9bee84d
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix message body processing when transitioning through the send flow.
Previously, the message text was largely unprocessed, but it wasn't
obvious like it is now with 980px viewports.
Also move 'table-cell' wrapping from inside the message template to Java
code so the in-place body replacement that happens during send can also
use it.
Bug: 13079016
Change-Id: I9a6a801c43b76c1714dfd2c8b41649fa6b9e90c8
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The side margin was in CSS px, but goalWidth was in dp. This caused the
goalWidth to be smaller than it should, which caused scrollWidth to be
smaller, which caused each message's zoom factor to be larger, which
caused text to be larger than it used to be.
Now. goalWidth and documentWidth both have body margins accounted for in
the proper units.
Bug: 12579959
Change-Id: Ice36b48f381cf3171b092ba27152aa56b64f4100
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Expand all conversations to 980px width to allow double-tap to continue
to work on newer Chromium builds. See
https://codereview.chromium.org/18850005.
cool side effects
* much more usable double-tap zoom factor on KK+ (now "it just works")
* better text rendering fidelity in complex layouts (text no longer
escapes bounds as it tended to do before)
* fixes text sizing edge cases where text was unexpectedly large and/or
line heights messed with text rendering
Initially enable these changes on both JB- and KK+ WebViews. Disable
NARROW_COLUMNS layout on JB- to prevent very short columns.
pros on JB-: better rendering fidelity like on KK+, consistency with KK+
cons on JB-: some power users will miss pinch-then-double-tap-to-reflow
functionality
Possible future improvement: use device-width viewport when full convo
fits to inherit Chromium's new fast-click-handling behavior.
Bug: 12579959
Bug: 10695551
Change-Id: I13f85a6df909a966fcd0862e5bd292ec6ae77212
|
|
|
|
| |
Change-Id: I9934b8b20734e17ede4060dfbb57f87ed6c89aef
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Uses the existing javascript image src rewriting
step to build a mapping of urls to message ids.
This mapping is passed into the InlineAttachmentViewIntentBuilder
so that it can use this information along with the
conversation id and account name to build attachment
and attachment list uris so that a photo viewer open
intent can be created.
Additionally, SecureConversationViewController will need a different
mechanism for this UI to work in Email/the eml viewer.
Change-Id: If14800348fe2191d0633bf768b8cb4e9746f6578
|
|
|
|
| |
Change-Id: I82cca1daa5c3fae375e72f24c76b68e44517e0f5
|
|
|
|
|
|
|
| |
Doesn't print attachment thumbnails for
image attachments. b/10712542
Change-Id: I5202338f60ddeb35238fce9d79691ba3c7a29e89
|
|
|
|
|
|
| |
Sometimes the other heights are too big.
Change-Id: Ib60321a0afece97f4024765f8682f8fca7c1e484
|
|
|
|
|
|
|
|
|
|
|
|
| |
We were occasionally using the wrong
height of the html document to determine where to place
the bottom footer. We're now using the proper place for the
bottom of the webview. However, there are still some
minor issues (a little bit of content being cut off
due to some other improper use of offsetHeight).
The fix for that issue will be tracked in b/10410912.
Change-Id: I2f6dbd46f52db3809e14ddb3e3257ebbc0fbe0b4
|
|
|
|
|
|
|
|
| |
Borders now expand and collapsed based upon the state
of the messages above and below them. Also fixed a bug
where replying to a message added an additional border.
Change-Id: Ia90caf4ea595767a90213fe33b29e1bd75c0aca0
|
|
|
|
|
|
|
|
|
|
| |
Previously we weren't adding the right items.
Need to not add a top border but add a border
at the end. Additionally needed to change the
javascript to also add the spacer for the bottom
border.
Change-Id: Ief82dca25c0c11b258cfe9a3d783d6c4f2e62974
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix Twitter emails by overriding any 'table-layout' style when
overriding table width to 'auto'.
Add infrastructure to blacklist transforms by sender domain.
Not actually used yet, because the Twitter email fix above is more
generally useful and doesn't need to be special-cased.
Bug: 8403536
Bug: 8685326
Change-Id: I9e72e8e6336c138674839724df57eef1f86a2d9b
|
|\
| |
| |
| |
| |
| |
| | |
jb-mr2-dev
* commit '6a507c356cd64a3b289c2b51d0294e243c520b36':
disable debug printing of message transforms
|
| |
| |
| |
| |
| | |
Bug: 8576229
Change-Id: I61d515360242c63f1cb51abc5c15eb882a916166
|
|/
|
|
|
|
|
|
| |
The <td> width transform is somewhat risky, so if we reach the end of
all transforms and the <td> change didn't help at all, undo it.
Bug: 8676694
Change-Id: I1767855483b1c2a0c78defc53a1fc5ea8fdccb93
|
|
|
|
|
|
|
| |
Also print it out in the message body in tiny text when the transforms
succeeded.
Change-Id: I942f707b5f7251f7aa9043c48fc0aa794c753065
|
|
|
|
|
|
|
|
|
|
|
| |
Improve the div width->max-width rule:
* apply it to <textarea>s to fix MOMA internal news emails
* override any min-width value
* switch from auto-width to 100% for empty textareas
* implement undo of this rule
Bug: 7400516
Change-Id: Ie72f7f80e9da8eb4c1a169e9b2bf56cf8a01c098
|
|
|
|
|
|
|
|
| |
Reapplying them was causing them to sometimes be backed out upon image
load, leading to text growing and shrinking several times.
Bug: 7400516
Change-Id: I5543361fd0aa60e233c27f28d243fbf037cbe1da
|
|
|
|
|
|
|
|
|
| |
It kicked in too often to warrant fixing 1% of unbreakable long email
signatures. Maybe we could bring it back if it could be properly scoped,
but I haven't figured out a good scoping rule yet.
Bug: 7400516
Change-Id: Id15ca8ec45783182d2c0fedd211a9b449ad12bb0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Shrink any <img> elements with an effective width wider than the
document. Instead of setting the width outright to the document
width, use "100%" + max-width so it can be smaller if the layout
allows.
Genericize the actionLog to be a list of functions, contexts, and
parameters. I would normally use bind() but A) it may not be available
on older WebKit builds, and B) it introduces an unnecessary closure.
Bug: 7400516
Change-Id: I5ab7f81674f9e238e6dc5ea04669d282c0586066
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Reverse <table> and <td> transform work when the resulting width is
unchanged, or significantly unchanged (below a threshold).
Switch <table> and <td> transforms to use CSS !important overriding
instead of modifying the elements inline. This is easier to reverse.
Add another <td> transform to aggressively trigger word wrapping intra-
word. This helps break long signatures, but it can render complex
promotional email unreadable. I'll have to hide this behind a classifier
that prevents this from running on complex messages.
Bug: 7400516
Change-Id: Ia17bae2b28eeec9d1cd40d0292380ae5d660586d
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Modify HTML content that is too wide if the experimental flag
is on.
Find <table> elements with a width wider than the body and
remove the widths in the hope that the table would otherwise
fit. Content authors often use this trick to prevent their
content from growing too wide in a very wide desktop browser
window.
Failing that, find all <div> elements where a width is
specified that is wider than the body, and turn the width into
a max-width.
Bug: 7400516
Change-Id: Ie4d3c2c4c50a191c30ee590dce71bc9bf6de8268
|
|
|
|
|
|
|
| |
Not very elegant, but it seems to help on Nexus 10.
Bug: 7400516
Change-Id: Iea7a67da63bb01bb58aad9c7b8035c90b33230e0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Adjust scroll position after a zoom to counteract jumping.
There's still a momentary flash, so it's not perfect yet. That
might be suppressable with ingenuity (i.e. gross hacks).
Suppress WebView input handling for an input sequence on
POINTER_DOWN to prevent WebView from scrolling during a scale.
I've chosen to implement message translation-during-scale
manually, so the WebView must not move on its own during a
scale, or else the translation will be doubly applied.
Bug: 7400516
Change-Id: I54c347c37d28864441c8a3bca3395bdc86f6616f
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
length for subject.
Also, a fix for:
b/7937635 Senders text displayed as "e": need to figure out why the single sender was being
marked as elided at all so leaving the issue open
Considering an option in which I divide the subject into 2 textviews and ellipsize based on
length...
Part of matching redlines.
Change-Id: I109001fb1d52c032d489bf7d3a0a6e925a9d7f1a
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When Auto-fit is off, disable built-in WebView zooming, which
is only capable of zooming an entire document (the entire
conversation).
Use ScaleGestureDetector to trigger a CSS 3D transform to zoom
in/out just a single message div.
During the gesture, the elements above and below the message
being scaled are untouched. This is ugly.
TODO: they should either move away (tricky) or at least fade
out.
When the gesture is complete, to avoid leaving overlapping
elements, we force-scale the height and reset the transform
origin to the top left.
TODO: auto-scroll to the correct place to counteract this
perceived jump.
Double-tap is not yet implemented. We'll have to do it
ourselves.
Bug: 7478834
Change-Id: I114e4977304c7060d499d116cc75bc0488967448
|
|
|
|
|
|
|
|
|
| |
The always allow pictures button only affects a single message in the
thread, so we need to propogate it to all other messages from the
same sender.
Bug: 7346217
Change-Id: I722ef284afd7b8853ddab622f54b86ebaaa5d463
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Towards the end of MR1, we saw more native crashes on
conversation load. It's possible it was purely due to increased
testing, but just in case it was due to loading animation
changes, go back to Gmail1-style content-ready code.
The document has a small div that initially has an
"initial-load" class set on JB+. The class is no longer
dynamically added in JS. The animation-start event listener
remains as-is, configured on DOM parse (and not later upon DOM
load).
Give the animation a different name from the CSS class that
triggers it, on the off-chance that this is crossing some wires.
Bug: 7370125
Change-Id: I4f0f807fbb9e32e024b6c74a6d9684e7fb337695
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Two-pane layout runs a stray layout pass upon send (related to
b/7409762). Prevent it from changing existing header positions
until the new body geometry is determined.
Fixing that exposed a separate issue, that conversations shorter
than one screen tall would not draw the new header, ever. Add
manual position refresh calls to newer JS calls that had
previously depended on ContentSizeChangeListener.
Bug: 7404104
Change-Id: I805fc9ab12878a95573316763986d84f89cfd2c0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Content within a super-collapsed block that was lazy-loaded
wasn't being processed for images and quoted text like initially
expanded content. Most images were caught by the regexp anyway,
but we need this step for completeness, and so quoted text
isn't out of control.
Patch up other cases of content replacement to hide images as
well.
Bug: 7392191
Change-Id: Ib784ce52c9c477635a125f261a2b9717b20887d2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Rather than rendering the conversation from scratch, look for
the special case of a single new message from the user that is
the last message, and just append the message HTML at the end of
the existing conversation.
To avoid stacking the new header on top of the previously last
message, do not notify ConversationContainer of dataset change
in this process. (An adapter dataset change means the container
will toss out all overlay views and immediately render given
the current spacers and adapter items. Do not want.)
The overlays for the new message will be drawn later when JS
notifies of a geometry change.
Bug: 7378693
Change-Id: Ifd2d3da8c57f685bbeae18f3d42c60821a97368c
|
|
|
|
|
|
|
|
|
|
|
|
| |
Disable zoom if all initially expanded messages are empty
looking-- that is, if their whitespace-trimmed text content is
zero-length.
Begin to dynamically assemble the meta viewport tag, which will
be useful for other advanced zoom-related tweaks.
Bug: 7340665
Change-Id: I86112617c18013e74b38af06d8a058aad315e9ec
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
and wire it all up to a pref that re-renders upon change.
separately, make a few improvements to whitespace management
when zoomed in with normalized mode:
* place attachment overlays at the top of their regions
* place overlays above the first expanded message at the top of
the conversation, instead of the usual position just above the
message.
* impose a max zoom factor to limit excessive spacer whitespace
Bug: 7312540
Change-Id: Iae3afff0ee81e4ba9367568e884a041780b24ebf
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When a reply is sent and synced, the quoted text portion
transitions from an expanded snippet to a collapsed one,
which shrinks the conversation height and causes the scroll
position to jump.
It's not enough to just avoid processing reply bodies, because
a reply to a wide message must also be normalized if the quoted
text portion is visible.
So when replacing a message body, switch the initial state of
quoted text blocks from collapsed to expanded.
Bug: 7246054
Change-Id: I5375beeb4209779e7cb7feaa570590a9cbc520da
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Gmail1 started the content-ready animation right away; the
initial state of the HTML had the animation already. Gmail2 was
waiting for the DOM load event before even starting the
animation, wasting a lot of time.
Starting the animation right away had a race condition where the
listener wasn't installed until later, so avoid any such race
by starting the animation immediately after registering the
listener.
Bug: 7258363
Change-Id: I00fc06998a06e1e4dce0d490ead9fbc6fef63ae0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Refine the image onload handling introduced in Idc1e1f21.
It turns out that resizing is pretty slow, and image onload
happens pretty often for threads or messages with lots of images
(pretty common). Rather than immediately handle onload events,
process affected Elements in a deferred handler to coalesce
consecutive onload events. This will also debounce situations
where a single message contains many images, as only one measure
and resize should happen.
Add timing logs to better diagnose conversation load issues.
Bug: 7278331
Change-Id: I252cee9c69955d993aea20379fcdc0beb973e372
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Save scroll position as a fraction of total height to handle
scaled and width-change cases. Restore scroll position fraction
in JavaScript as soon as the document geometry is known.
(This was how it was done in Gmail1, and I can't find a better
non-JS way to detect this signal.)
Does not restore scale upon rotate/re-render.
Bug: 7246054
Bug: 7022509
Change-Id: Ibf811b31ba6d7d79d34c236f6ffe2229d626528f
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Make several improvements that, together, prevent the
conversation WebView from zooming out when revealing
wide message content.
Disable WebView's overview mode. It wasn't doing anything
special due to our viewport meta tag. Leaving it enabled
caused newly revealed very-wide content (beyond the reach of
CSS zoom: see b/6053074) to zoom out the entire conversation.
Taking it out keeps the scale level at 100% unless the user
zooms.
Shrink newly revealed content. Message expansion,
"Show quoted text", and image onload can all reveal new wide
content. Revealing repeatedly will not repeatedly shrink.
Quoted text will only shrink the quoted portion, not the entire
message.
Finally, override default handling of <pre> blocks to match
desktop Gmail. This obviates the need to zoom out at all for
messages like build notifications and other programatically
generated mail.
Bug: 7215221
Change-Id: Idc1e1f211aa3463a68146bef2846583cd7d075d3
|
|
|
|
|
| |
Bug: 7154554
Change-Id: Iacdf4bb7466978c0da3e3985d14e42894dac2ff7
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Sending a new message updates sender string (for custom froms
and such) and the body (for elided text), after the message is
sent and synced. Avoid conversation view flashing (re-render)
in those cases by applying the changes in-place.
Sender string changes will update all dependent adapter items,
and a dataset change on the adapter re-renders all overlay
views.
Body changes are handled by sending affected IDs to JS so it can
request individual message bodies. Quoted text collapsing is
then re-applied on each new body.
Bug: 7183109
Change-Id: I630187c276ffc63448e6b457c9cd02afd452c438
|
|
|
|
|
|
|
| |
Shows content faster
Has the background white-screened out until we get the signal
part of fix for b/6268986 Perf regression (~5 seconds) in loading messages on orientation change in new gmail
Change-Id: I653c40c88f0cb5cd0e31655ad89762c6f07dff66
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
fixes b/6272925 jank - conversation loading isn't animated
fixes b/7131797 Blank scren shown for live conversation, when in airplane mode
Note: need to do a next cl that puts off showing the spinner
until at least XXX ms have passed
Also, we cant use the software layer for rendering like we did
in gmail1
it looks like there is a skia crash that triggers when we try
to use software / switch to hardware rendering for the browser
need to investigate that more
Change-Id: I96a30b700c3e88d52e603fe2f11b44d113e013d2
|