Mullvad

Recent Commits

• Markus Pettersson(03 Sept 26)

  Merge branch 'fix-node-cves'

• Markus Pettersson(03 Sept 26)

  Bump `fast-uri`

• Markus Pettersson(03 Sept 26)

  Bump `browserslist`

• Markus Pettersson(03 Sept 26)

  Bump `@xmldom/xmldom`

• Markus Pettersson(03 Sept 26)

  Bump `@humanfs/node`

• David Lönnhager(03 Sept 26)

  Merge branch 'win-extend-dns-flush'

• David Lönnhager(02 Sept 26)

  Extend DNS cache flush timeout Mitigates timeouts likely related to large hosts files

• David Lönnhager(02 Sept 26)

  Do not flush DNS cache on DNS reset We can assume that the cache isn't poisoned when there's a tunnel

• Kalle Lindström(03 Sept 26)

  Merge branch 'make-unsupported-info-less-intrusive-for-alpha-and-dev-droid-2706'

• Kalle Lindström(02 Sept 26)

  Do not show version notification for alpha and dev

• Kalle Lindström(02 Sept 26)

  Use different icons for alpha, dev and release

• Jonatan Rhodin(03 Sept 26)

  Merge branch 'fix-kotlin-dependency-not-being-pinned'

• Jonatan Rhodin(03 Sept 26)

  Fix some kotlin dependencies using the latest version of kotlin

• Linus(03 Sept 26)

  Merge branch 'ios-rust-ffi-single-job'

• Linus(03 Sept 26)

  Build and lint in one job and test the iOS FFI crates in another Clippy reuses the target directory the build just populated, so keeping the two in one job measurably shortens the lint. Measured over 30 runs of the old layout the lint took 2.91 min from cold against 1.80 min after a build in the same job. The tests build for the host rather than the iOS target, so they shared nothing with the build and are better off in their own job, where a red check says whether the tests or the build and lint failed. Clippy is split per feature set for the same reason.

• Linus(03 Sept 26)

  Remove the iOS simulator target from the Rust FFI workflow The simulator target differs from the device target only in target_abi and target_env, and in the simulator's wider target_feature baseline. No crate in this repo branches on any of those, and both targets resolve to an identical dependency closure, so building and linting the simulator caught nothing the device target did not. With one target left the matrix goes away. The target name moves to a job level env variable so it is still stated once per job.

• Linus(03 Sept 26)

  Set the iOS FFI package list and shell in one place The two crates were named in every cargo invocation, so adding or renaming one meant editing each in step. Substituting the list also shortens the clippy commands enough to fit on one line. Two steps declared `shell: bash` while the rest relied on the default. That is not only cosmetic: an unspecified shell runs `bash -e`, whereas `shell: bash` runs `bash --noprofile --norc -eo pipefail`, so only the explicit form fails a step when a command in a pipeline fails.

• Jon Petersson(03 Sept 26)

  Merge branch 'devices-logs-are-duplicated-when-being-sent-ios-1767'

• Jon Petersson(01 Sept 26)

  Remove duplication of logs

• David Lönnhager(02 Sept 26)

  Log failure to compile shaders instead of showing alert

• David Lönnhager(02 Sept 26)

  Merge branch 'bound-driver-removal-timeouts'

• David Lönnhager(02 Sept 26)

  Handle unpopped values from nsExec::Exec in Windows installer The intended "return value" was passed in as an argument instead. Likely caused no issues in practice other than growing the stack for no reason.

• David Lönnhager(02 Sept 26)

  Kill mullvad-setup.exe if it lingers after timed out device cleanup

• David Lönnhager(02 Sept 26)

  Update changelog for removed Wintun adapter cleanup

• David Lönnhager(01 Sept 26)

  Get rid of RemoveAbandonedWintunAdapter The macro was added in 2021.3-beta1 to clear an adapter holding our fixed GUID even when it belonged to a different Wintun pool (#2505). "Pools" have not been relevant since 2022.5-beta1. Wintun dropped pools in 41a962e7d0.

• David Lönnhager(01 Sept 26)

  Do not loop forever when a device removal needs a reboot

• David Lönnhager(01 Sept 26)

  Add timeout for removal of abandoned drivers in installer From a log: [13:04:19.377] RemoveAbandonedWintunAdapter() [13:37:46.381] RemoveAbandonedWintunAdapter() completed successfully Since these cleanups are nice to have rather than strictly necessary, this commit adds a timeout.

• Emīls(02 Sept 26)

  Merge branch 'ios-gotatun-3-path'

• Emīls(31 Aug 26)

  Leave TODO to remove soon to be dead code

• Emīls(13 Aug 26)

  Test GotaTun reachability handling Covers offline blocking and adapter teardown, starting while offline, restoration reconnects, socket recycling on satisfied paths, debounce of repeated unsatisfied updates, and observer lifecycle. DefaultPathObserverFake previously swallowed updatePath calls; it now delivers them, delivers the current path on start like real observers, and gains a GotaTun counterpart.