4% of our crashes are from accessing latestContactsById with an invalid
address. This was partially addressed by assigning a value on init, but
it's still happening.
By converting the ivar to an atomic property we can avoid any funny
business that would happen when accessing this var while it's being
updated in a different thread.
Also, make sure we're copying to defensively strip mutability.
Also:
* bumped release target
* removed unused code
* dealloc at the top per style.
* Rebased OWS fork of SocketRocket against upstream
* Pulling in SocketRocket directly from OWS Github rather than going
through the motions of publishing a pod that only we use.
* Bump version
// FREEBIE
While trying to pull in upstream changes to our fork of SocketRocket I
noticed we're publishing it/installing it from CocoaPods trunk, but
there's no need to do that.
Also, due to upstream refactoring, the actual included file has
changed.
// FREEBIE
* CocoaPods 1.0 compatibility. Thanks to @brianherman and @orta for the Podfile changes, taken from PR #1156
* Modified Travis-CI to use the newer CocoaPods 1.0 as well.
* Performance improvements to Travis-CI to reduce build time by 50%-75%
Cherry-picked and conflicts resolved by @michaelkirk
* bump version
* pulled latest translations
* Register localizations with app.
- Persian (fa)
- Indonesian (id)
- Macedonian (Macedonia) (mk)
- Shona (sn)
- Galacian (gl)
(Project > Signal > Localizations > Add Localizations)
Generalized Macedonian locale since it seems like iOS doesn't support
the "Macedonia(Macedonia)" sub-locale
* disable farsi until proper RTL interface work is done
// FREEBIE
* modify script to fetch all uploaded translations.
Otherwise we don't always get the latest translations.
This might be because we're only fetching "reviewed" translations, but
so-far-so-good with unreviewed translations, and we don't have a real
"reviewing" hierachy in place anyway. This policy might change in the
future, but at least now it's not ambiguous.
// FREEBIE
* latest translations
* Fixes "New Message" --> "Invite contact" exception for iPhone devices that do not support SMS messaging
* fix SMS invite on empty inbox
This code appears in two contexts - (1) in an empty inbox , where no other view
controller is presented, and (2) in compose interface where a
search controller is presented, and must first be dismissed.
* only check SMS sending ability directly, (not device model)
This is better because:
1. sometimes iPhones can't send SMS
2. Sometimes iPads/iPods can send SMS
* correct localization key
* bump build
// FREEBIE
* all Signal users can send text messages, never hide texting toolbar.
//FREEBIE
* Fix composition box size when switching threads.
Partial revert of 2c83046ff6 which
introduced a shared reusable message view controller across threads.
2c83046ff6 resulted in several discovered
bugs so far (#1179, #1150, #1152, and maybe: #1146). It's pretty clear
at this point we're going against the grain of how
JSQMesageViewController is intended to be used, and since the nominal
purpose of this feature (iPad Layout) doesn't exist, we should revert to
the known good way of interacting with the MessageViewController,
creating a fresh instance per thread.
// FREEBIE
* bump release
//FREEBIE
* fix invite over sms alert
Could not present the sms invite alert controller because we were
already presenting the UISearchController
FIXES: #1182
// FREEBIE
* Update translations (4 new translations!)
*New Translations*
- Persian (fa) (100%)
- Indonesian (id) (82%)
- Macedonian (Macedonia) (mk_MK) 100%
- Shona (sn) 98%
Inluded a script to pull translations that are >=80% complete
80% is completely arbitrary, but nicely bisects the current translation
efforts which fall into buckets of "over 80%" and "under 50%."
// FREEBIE
* final tx pull before push source
//FREEBIE
* tx pull, after updating source (plus some eager beavers already pushed new translations)
iStringVerify passes
* Make "add person" label more i18n flexible
This accommodates (e.g.) Italian, which was being truncated.
// FREEBIE
* Galician localization
Moitas grazas xesusmosquera!
// FREEBIE
* i18n for Unarchive action
started leaving hints for our translators. These are currently manually
redundantly duplicated in the source/strings file, in hopes of a more
automated future with genstrings.
// FREEBIE
* include iStringsVerify check in update-translations script
and pull latest translations after including UNARCHIVE action
// FREEBIE
* i18n GROUP_MEMBERS_HEADER
//FREEBIE
* pull translations
* Group actions bar button image was too big
Added image inset to resize.
// FREEBIE
* return blue line to neutral gray. (partial revert of 0c1a97a743)
// FREEBIE
* Podfile.lock corollary to updating SSL in b7813bdc9a
// FREEBIE
Since we're now re-using the message controller across threads, we have
to reset some elements to their initial state when switching threads.
Missed this one.
Fixes: #1150
// FREEBIE
* Fix ability to attach photos from camera
Looks like this came about with the animated GIF handling. We'll only
go down the byte-comparison-mime-type-detecting code path for attaching
existing photos, since it only exists for animated GIFs.
This will also revert to properly compressing our image attachments, so
long as they are taken from the camera.
* Prevent crash when tapping broken image
It never makes sense to present a "full screen" nil image. Previously
this happened when camera-capture was broken, but could conceivably
happen for other as-of-yet unknown reasons.
// FREEBIE
*most* of our time when scrolling around a message view is in this
method. Doubly so when these are media items.
remove interaction from cache when it is deleted/updated
// FREEBIE
* fix thread preview when deleting last message
because we are doing internal accounting of what our "last" message id
is, when we delete the last message, other things get out of whack.
Caching is hard, don't do it! We can efficiently get the last object
using our existing Yap DB view.
//FREEBIE
* remove lastMessageId in favor of lastMessage for cleaner code
//FREEBIE
Since "Key Path" is the header for run time attributes, my guess is that
this was a click misfire so long ago. I'm not clear if this was ever
causing a crash or just a warning, but we don't want the noise.
Fix "Unbalanced calls to begin/end appearance transitions"
You should not segue while animations are in progress. I'm not sure if
this was causing intermittent crashes or just the warning.
//FREEBIE
Closes#1130