This change modifies the reported version number for km-config:
if km-config is installed from a debian package, the package
version will be output as well, e.g.:
`km-config version 14.0.252-beta (package version 14.0.252-1+bionic1)`
Otherwise the output will be identical to before:
`km-config version 14.0.252-beta-local`.
This change also modifies the script that sets the version number so
that builds on Jenkins are now detected in addition to TC.
This fixes#4579.
Allows us to manually run a version increment, for example if there is
a problem with a release build such as running out of space on the
download server.
Fixes#4484.
tsysinfox64 was being signed, but too late in the build
process as it is embedded into tsysinfo.exe. There was
code to sign it earlier, but that was being skipped due
to a missing build flag. This has been a problem for
some time.
This is necessary to be able to distinguish between a newly triggered
tagged build and a retriggered tagged build. The first tag param gets
saved between builds, but then we need a second parameter that doesn't
get saved that tells us if this build is a retriggered one.
Fixes#4201.
This adds a comment to pull requests with information about which build
a particular PR will be landing in, based on the HISTORY.md we have
already updated.
Fixes#4061.
Four things:
1. Reorder the script steps so that builds are triggered before the PR
is created. This should not really be necessary, but I am seeing some
evidence that TeamCity is not respecting the commit sha we request to
build. (This arose with 14.0.194 where a build agent became
unavailable during a build and TC cancelled and restarted the build,
but from a later revision... which this particular fix may not help,
I guess...)
2. Add debug logging so we can see exactly which commit sha that we
send to TeamCity
3. Handle errors in version increment by aborting the script with an
error.
4. Add a -f force mode which we can use to trigger a build even if no
changes are detected. This will need a corresponding patch in TC.
Instead of just reporting on --first-parent, now lists all merges that
are in the log, which means we get some non-PR merges but that's ok. At
least we get all of them.
This was most visible with chained PRs which were merged bottom-up.
The background reason for this is to test a fix to the web build, where
s.keyman.com resources were not being updated since we refactored the
build configurations (sadly, the build did not die when it should have).
When displaying the Info page of alpha builds from the Play Store, we noticed the version string displayed the wrong environment:
`14.0.X-alpha-local` (CI builds should have ended with `-alpha`).
It turns out the `VERSION_` shell variables from `build-utils.sh` aren't visible in `version.gradle` (though CI environment variables like `build_counter` are visible).
This PR:
* Exports the two VERSION strings needed in version.gradle
* Updates the Info page to simply display `BuildConfig.VERSION_NAME`
Follow-on to #3300
* Update the FV Android app to use the same build script to download its default keyboard package. (fv_all.kmp)
* This means downloading fv_all.kmp is the only valid option for the build (remove option to copy/rebuild keyboards)
* Reduce the number of exports in the Bash scripts
* Also removes sil_euro_latin.kmp since the app will fallback to fv_all/sil_euro_latin (in a future PR)
