| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
* PROBLEM: the build fails on gitlab in a debian-based docker container
* BUT: i (@aguestuser) have a recently-achieved passing build on a
debian laptop
* ATTEMPTED SOLUTION: construct a dockerfile that matches my local
configuration as precisely as possible
* PROGRESS: the build gets further than it did before -- getting part of
the way through the `buildNative` gradle script before failing
* REMAINING FAILURE: several arm64 cross-compile steps in the
`ndk-build` step fail because they depend on
[neon](https://developer.android.com/ndk/guides/cpu-arm-neon.html):
```shell
[arm64-v8a] Compile : crypto_static <= aesv8-armx-64.S
openssl/crypto/aes/asm/aesv8-armx-64.S:35:2: error: instruction requires: neon
eor v0.16b,v0.16b,v0.16b
^
openssl/crypto/aes/asm/aesv8-armx-64.S:36:2: error: instruction requires: neon
ld1 {v3.16b},[x0],#16
^
openssl/crypto/aes/asm/aesv8-armx-64.S:38:2: error: instruction requires: neon
ld1 {v1.4s,v2.4s},[x3],#32
```
* PROPOSED NEXT STEPS:
* consult team to see if there's any collective wisdom about `neon`
* look for ways to analyze diff of c dependencies in local machine
v. docker instance
* consider using ubuntu or debian:sid as the base image for the
android container?
|
|
* add detailed instructions on installing and compiling
* provide workarounds for gotchas setting up emulator on debian
* offer instructions for building in docker
side-effects:
* update build tools & gradle version
* fix indentation in build.gradle
* comment out tests in `TestLogFileHandler` causing `build` to break (and provide justification)
|
|
closes #8536
|
|
0.9.5RC2
|
|
remove pins, because didn\'t work this way (solves #8455)
|
|
ics-openvpn already shows it if necessary. Our heuristic (just looking
for an "error" keyword in the past N messages of the log) is very weak,
and it returns an annoying false positive: turning off the VPN triggers
the show log error.
|
|
- Android gradle plugin to 1.5.0
- Gson to 2.4
- Support annotations to 23.2.1
- Robotium-solo to 5.5.4
|
|
onCreate gets called when the activity is back on the screen. If the app
was already launched, we should not set everything up again.
|
|
|
|
|
|
|
|
Warn about the needed libraries for 64 bit systems.
|
|
This is also an open issue in Transifex: https://www.transifex.com/otf/bitmask-android/translate/#es/$/22692144?issue=yes You should double check if this is correct.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Riseup's EE certificate expired (EE = End-Entity, the commercial
certificate signed by the issuer), and we weren't able to fetch
provider.json because the pin wasn't valid.
This problem needs to be avoided in the future, using
HKPK (https://tools.ietf.org/html/rfc7469), which Micah's implementing
on Riseup.
Switching from build types from flavors enables us to run tests against
production apk, not just debug. I didn't detect this pinning problem
because tests were run only against the debug apk, which trusted
preseeded providers by default (thus bypassing pinning issue).
|
|
|
|
|
|
|
|
|
|
ics-openvpn is now officially on GitHub, and they track openssl and
openvpn as submodules, so it's easier to update everything. Just a git
submodule update --recursive.
I've also set up soft links to native modules from ics-openvpn in app,
so that we don't copy files in Gradle (which was causing problems with
the submodules .git* files, not being copied). That makes the repo
cleaner.
|
|
|
|
|
|
|
|
Sometimes, the emulator is so slow (I wish genymotion was open
source...) that even though everything's going OK tests fail because
they don't wait enough.
|
|
testVpnButtonIsDisplayed and testVpnIconIsDisplayed fail because calyx in testVpnEveryProvider fails, but if you run them alone they pass.
|
|
Give more time to provider details fragment to appear, because debugging makes the code run slowly.
|
|
Updated butterknife, dagger, renderers, fabbuton and gson, as well as
gradle plugin.
|
|
The FabButton doesn't scale the icon as big as I want it to be, and it
doesn't let me set the size to "wrap_content" because the library sets
the sizes to match_parent.
I'm going to try to modify my fork of FabButton to see if I can obtain
what I want in its demo: just an icon and the progress indicator, the
icon as big as possible.
|
|
And checks in controllers, so that if a button isn't shown, I throw a
new IllegalStateException. This helps to trace the error.
|
|
|
|
|
|
This way, I can add more easily the tests for errors.
|
|
|
|
Next step: beautify.
|
|
|
|
Tests click new button, instead of action bar's.
Fixing more bugs...
|
