This is done by the build_ubuntu* jobs.

This makes it more consistent.

The CI fails with
go: downloading go1.22.2 (linux/amd64)
go: download go1.22.2 for linux/amd64: toolchain not available

this updates the jobs that builds the client on windows and
macOS on github actions

the jobs were failing due to missing qt-installer-framework
as well as failing to properly install qt6 on windows these
jobs now uses `aqt` to install qt6 on both win and macos

the backend.quit emits the 'quitDone' signal which also
triggers the 'aboutToQuit' signal for the  QApplication
since its handler calls 'QApplication::quit()'

the handler for 'aboutToQuit' signal already calls Quit
so removing it from backend.quit avoids another call to
Quit

This reverts commit b9fc3584.

the aboutToQuit signal is emitted just before the app closes and
gives chance to run some teardown routines, when closing the app
via a keyboard shortcut (cmd+q on macOS) we get this signal  and
the handler stops the firewall and performs other needed cleanup

removing this causes a bug where quiting the app would keep  the
firewall running which results into no internet access

since obfs4 and kcp is disabled in the UI when provider
is riseup, transport needs to be set to openvpn to work
around config file with Obfs4 and KCP set to true  from
a previous version of the app

riseup currently is not having any deployed obfs4 or kcp bridges
but the eip-service.json contains transports[].type as obfs4 for
many of the gateways, therfore the bridge options are explicitly
disabled if the provider name is 'riseup'

this prevents starting openvpn connection automatically after
app start, since currently there's no preference setting  for
it to allow the user to control the behavior

this also sets DisableAutostart to true since it is currently
not working on any of the platforms

this uses the contentItem property of RoundButton to
use an Image item and set the power button as on/off
changing from Icon to Image allows to use the mipmap
property, which should improve the image quality

the Stop is blocking and cannot be called concurrently
with Start, when for some reason obfs4 client fails to
connect and has not yet returned from Start, a call to
Stop will block until Start has returned, which  makes
UI unresponsive for some time

calling it in a separate Go routine works around  this
issue as we can proceed to stop the openvpn process

jkito authored 7 months ago and jkito committed 7 months ago

in case of KCP the transport.Type is still obfs4, but it has
a different port than obfs4, without this commit KCP enabled
bridges were also considered as valid obfs4 bridges

in theory the same bridge can be running in both obfs4 and KCP
mode and listening on different ports, but the eip-service.json
schema separates this into different transport objects

this fixes a bug where after selecting obfs4 transport it was
trying connect to a KCP enabled bridge with plain obfs4 mode

jkito authored 7 months ago and jkito committed 7 months ago

during startVPN any error occuring are set but
this was not done for stopVPN

kwadronaut authored 7 months ago and jkito committed 7 months ago

android, qt, tranxifex all have different ideas on how languages
and locales should pose themselves. It is not relevant for  most
languages where we only have one locale, like Turkish. But if we
have different ones or different scripts, we need to tell  while
downloading updates where to look for them.

this updates the internationalization doc with explanation about
transifex language mapping and removes duplicate language files

solves #888

kwadronaut authored 7 months ago and jkito committed 7 months ago

we currently have language entries which appears to be repeated
but infact they are different languages, e.g es_AR, es_CU  both
show up as 'Spanish' in the language picker, to make it  easier
to distinguish between them the region is added to the lanugage
picker list, so es_CU now appears as 'Spanish (Cuba)'

Pea Nut authored 8 months ago and jkito committed 7 months ago

Build -> Ubuntu -> Debian -> Arch
From most relevant to least.
The build instructions clarify the PROVIDER env, which is used later in
package_deb.

the post-install tool run by the installer needs to know the path
for the app bundle as the bitmask privilege helper is located  in
the app bundle, earlier it was using a glob pattern "*VPN.app" to
find it, but since in the previous commit the BitmaskVPN name was
dropped and now uses Bitmask instead, this glob pattern no longer
works, so the app name is passed as argument to the tool

this added metadata is used by accessibility tools for
aiding the user to interact with the application

since the application is changed the name is updated
in the README as well to keep the naming consistent

the android app is called as Bitmask, this renames the desktop
app to use the same name and keep the naming consistent across
platforms

this was missed in 37a714e7 where the value of kcp
was written to the config file but during  parsing
the value was not added to the local struct

in b1796790 the config.Provider was changed from the provider
URL to the provider name but the check fetching motd expected
an URL this updates the case statement to match the  provider
name instead of the provider URL

kwadronaut authored 8 months ago and jkito committed 8 months ago

this updates the qt linguist ts files to target the strings
correctly after the recent UI changes

this adds a new boolean "KCP" key to the systray.json
config file

kcp is linked to obfs4, so setting kcp to 'true' sets
obfs4 to also true