bleedingdeacons / link / 34550221772
92%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 19
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 95.266%. Remained the same
34550221772

push

github

web-flow 
perf: make the Android dev loop seven seconds instead of two minutes (#31)

* perf: make the Android dev loop seven seconds instead of two minutes

LinkAndroidOnly set TargetFrameworks to a one-entry list, and a project
with TargetFrameworks *plural* is a cross-targeting build however short
the list: MSBuild runs an outer build that dispatches to an inner one,
and the outer build has no `Run` target. That MSB4057 is why the loop
went through EmbedAssembliesIntoApk=true and a hand-rolled adb install
-- a 90 MB APK pushed over USB on every edit.

Setting TargetFramework singular instead makes it an ordinary inner
build. `Run` exists, so one command builds, installs and launches; and
because nothing hand-installs any more, Fast Deployment is the point
rather than the trap. The Debug APK is 14 MB.

Measured here: an edit to one source file reaching a running app went
from about 2m15s to 7s. Steady state with nothing changed is 22s cold
and 7s warm.

CI is unaffected and slightly faster. Its Debug build is compile-only
analysis, and its artifact is a Release build, which embeds assemblies
regardless -- verified locally, 21 MB and correct.

LinkIosOnly deliberately keeps TargetFrameworks. Nothing runs -t:Run
against an iPhone from here, and now that CI builds the iOS head only
when asked, a subtle break in that path would sit unnoticed.

* ci: let the ios label actually trigger a run

pull_request defaults to opened/synchronize/reopened, so the label the
iOS job reads changed nothing until somebody pushed -- and the run that
had already decided to skip the head did not reconsider. Asking for the
iOS build meant making a commit to ask.

Any label change now starts a run, because types cannot be filtered by
label name and the filtering has to stay in the job's if. PR runs share
a concurrency group with cancel-in-progress, so that is bounded.

1127 of 1183 relevant lines covered (95.27%)

15.81 hits per line

Jobs
ID Job ID Ran Files Coverage
1 34550221772.1 19
95.27
GitHub Action Run
Source Files on build 34550221772