MapColonies / shigola / 5cac2df3c
44%

DEFAULT BRANCH: master
Ran
Jobs 2
Files 183
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: 44.353%. Remained the same
5cac2df3c

push

github

web-flow 
build: make the build cgo-free end to end (MAPCO-11489) (#14)

* build: stop enabling cgo in the devcontainer and the release builds

Nothing in this tree has been compiled conditionally on cgo since the
GeoPackage provider went (MAPCO-11488), but the build configuration went on
acting as though something were. The devcontainer set CGO_ENABLED=1 outright,
the release image installed build-base for "CGO support", and the Amazon Linux
action existed so a cgo-linked binary matched the Lambda runtime's glibc.

Worse, the release matrix never agreed with itself: with CGO_ENABLED unset,
linkage was decided by whether the runner had a C compiler and whether the
build was a cross-compile, so the arm64 artifact was already static while the
amd64 one was not. Pinning it to 0 in every release build makes the published
binaries the same shape as each other, and drops the build-base install that
cost the image build about 1:30.

The devcontainer's setting is removed rather than zeroed: `go test -race` links
a C runtime and so genuinely needs cgo, and that is a property of the test
command rather than of this tree.

TestNoCgoEnabledInBuildConfig keeps it that way. It is the configuration-side
counterpart to TestNoCgoConstraints: that one says nothing here is compiled
conditionally on cgo, this one says nothing here turns cgo on. Verified to fail
on each form it claims to cover -- YAML, a Dockerfile ENV and a shell
assignment -- by injecting one of each and watching all three be reported.

Refs MAPCO-11489

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* docs: document a single test mode

CONTRIBUTING asked for the suite to be run twice, once per CGO mode, and every
change since the GeoPackage provider was removed has paid that cost for a
dependency that no longer exists. One mode now, with the reason stated so the
rule is not reinstated by reflex, and a note that `-race` is why CGO_ENABLED is
left unset rather than pinned to 0.

The conformanc... (continued)

7889 of 17787 relevant lines covered (44.35%)

79.78 hits per line

Jobs
ID Job ID Ran Files Coverage
1 5cac2df3c.1 183
44.35
GitHub Action Run
2 5cac2df3c.2 183
44.35
GitHub Action Run
Source Files on build 5cac2df3c