There's some corner cases that were very badly captured. On the first
place, it's confusing to quit on connection errors.

Secondly, a side-effect of aborting the initialization of the bitmask
object was a semi-random segfault when trying to access the object.

Here I pass any connection errors to the gui, but leave to the user to
quit the app. This probably will need more work when we want to
terminate the app on unrecoverable errors (no polkit, etc...), but for
now it makes the systray much more usable.

- Resolves: #465

Kali Kaneko authored 4 years ago and meskio committed 4 years ago

a rather annoying bug, since it keeps popping up for every status
change. to be included in a hotfix for 0.21.2 release.

- Release: 0.21.2
- Closes: #459

forcing a style override looks like a good workaround to avoid system qt
plugins mixing up with the libs in the snap.

- Resolves: #458

10 seconds might be too low nowadays when accessed from some regions.

- Related: #463

Kali Kaneko authored 4 years ago and Kali Kaneko committed 4 years ago

Let's use openssl for the time being, there's a compilation issue with mbedtls
that I don't seem to be able to solve.

- Closes: #311, #453

kali authored 4 years ago and Kali Kaneko committed 4 years ago

stop if:

- bitmask is running (we don't want to mess with the helper if the app
  is running)
- we cannot stop the service for some reason.

powershell gives more flexibility to stop services, but Remove-Service
is not present in PS < 6, so we try to remove the service from the
post-install too (the golang helper will complain about an eventlog
registry key).

this could probably be improved in the helper, but we'll be moving to
the official openvpnserv2 service for 0.21.4. we might want to revisit
the current helper for the firewall/killswitch.

Kali Kaneko authored 4 years ago and Kali Kaneko committed 4 years ago

- install into global /Applications
- document how to troubleshoot helper
- uninstall app is visible on top-level folder
- improve detection of running processes for old and new binaries

- Closes: #441
- Closes: #445
- Closes: #435

it had been changed in the hook itself, but this script kept using the
deprecated form.

- Resolves: #431

- Closes: #436

we should try to use qmlfmt, indeed

- remove unused debug
- uncomment notification function, leftover from debug
- comment about go-land mutexes

Kali Kaneko authored 4 years ago and Kali Kaneko committed 4 years ago

Let's make it easier to tell us how buggy our software is, or how happy they
are to use it :)

- Closes: #422

Kali Kaneko authored 4 years ago and Kali Kaneko committed 4 years ago

this is a regression that was introduced with previous changes to the
menus in windows. activation segfaults in linux and osx, so capture the event
only for windows. did not investigate further about the causes, but it seems
like a good workaround for the time being.

Kali Kaneko authored 4 years ago and Kali Kaneko committed 4 years ago

This fixes the segfault.

mRootItem has to be deallocated to avoid memory leaks, but doing it
concurrently with the dump of the json was a terrible idea.

- Closes: #360

- Resolves: #413

They are hidden in transifex.

- Resolves: #414

decodestring is an alias that was deprecated in 3.1

- Resolves: #412