The backend returns them order by when the post was favourited;
reordering them by post date jumbles everything up each addition
and serves no purpose.
Fixes: AkkomaGang/akkoma-fe#391
This pulls in 267 new emoji:
- all 258 non-deprecated country or macro region
flags (composed by two regional indicators)
- all 3 regional flags currently recommended for general use
(Wales, Scotland, England)
- a few random ones i picked out
- goose
- heart on fire
- heart mending
- transgender flag
- rainbow flag
- pirate flag
The new names are derived from official Unicode names
with minor modifications to fit into the usual shortcode scheme
and dropping the flag_ prefix from country indicators.
Due to a naming conflict the old "japan" emoji
U+1F5FE SILHOUETTE OF JAPAN was renamed to "japan_silhouette".
Easy Japanses (ja_easy) and traditional Chinses (zh_Hant) use
(custom) non-ISO codes in the interface. Because MastoAPI only accepts
ISO 639 codes, the backend will return an error rendering users
unable to do anything unless the post’s language was explicitly set.
It was merged into pleroma-fe on 2022-02-03 in
76547fe66d and imported
into akkoma-fe on 2022-06-08 with the merge commit
f6cf509a04.
However, something went wrong in the merge and while the setting
and its infrastructure exist, it is never used anywhere and @ is
always displayed as text.
Given it existed in this broken state for nearly one and a half years,
never worked on akkoma-fe and no bugs were filed about this, it appears
nobody cares, so let’s just remove it.
Notifications about favourites and follows use .notification-right,
notifications about replies instead use .heading-right.
Previously only the former set a min-width, however the
chosen value of 3em was too small to fit the worst case.
As a consequence, when the timestamp text changes over time,
its element width changes, which may result in neighbouring text
(no longer) needing to wrap to a new line in turn changing the size
of the whole notification box pushing older notification boxes down/up.
These constant movements at the side of the screen can be quite
annoying and confusing when the cause cannot be immediately discerned.
Avoid this, by reserving enough space for any timestamp.
For English, the worst case is the five-character 'XXmin', since the
short identifier for minutes is the longest with three letters.
With two exceptions, all other current localisation also do not exceed
three letters in any short unit identifier up to days.
However, some localisations (e.g. Polish) additionally insert a space
between numerical value and unit. This matches SI recommendations
pushing the worst case to 6 characters.
6 characters will be sufficient for timestamps up to 3 weeks in all
languages (minus prev exceptions), which seems reasonable enough
as beyond this timestamps rarely change anyway.
The aforementioned exceptions being Vietnamese and Occitan,
but in the current localisation all or the relevant short unit
identifiers are identical to the long forms indicating this is
just due to incomplete translation.
Indeed, Vietnamese Wikipedia (read through machine translation) suggests
“ph” is commonly used as unit identifiers for minutes, but the current
localisation fully spells it out as “phút”.
Currently all notifications except follow-related once include
and explicit direction text. (It missing in follow notifs is due to an
omission in 804ba0cdb6 which only added
the newly introduced with-direction to status-related notifs. Before,
presumably all notifs included direction text.)
But in the notification tray horizontal space is scarce
and notifs can already be assumed to only come from the past.
While it might not be too bad for the English localisation’s 4-letter
' ago' suffix, e.g. the Indonesian localisation’s ' yang lalu' needs
10 letters.
Thus instead of fixing the omission for follow-related notifs,
drop direction text from all notification timestamps.
Modern browsers start to tighten down on third-party access to cookies.
E.g. in current Firefox, a warning about the userLanguage cookie was
shown since it did not yet explicitly set the SameSite attribute and the
default is about to change.
The cookie name being referred to as BACKEND_LANGUAGE_COOKIE_NAME
suggests it should be readable by the actual Akkoma backend, which can
live at a different domain than akkoma-fe. Thus explicitly enable
sharing with third-party sites.
No warnings were shown for other cookies, so I assume
this was the only one not yet setting SameSite.
