Source: glib2.0
Version: 2.58.2-1
Severity: serious
Justification: fails to build from source (but built successfully in the past)
A test fails on arm64:
START: /<<PKGBUILDDIR>>/debian/build/deb/tests/refcount/closures
init 0xaaaafde2c210
.
.
.
.
c.
.
.
.
c.
.
stopping
stopped
**
GLib:ERROR:../../../tests/refcount/closures.c:298:main: assertion failed:
(seen_thread2 != FALSE)
I've seen this before, in 2.54.2-3. If we run too many iterations, the
test times out; but if we run too few, the main thread starves the other
threads and the test gets through its intended number of iterations in
the main thread before the other threads have been scheduled at all, by
starving the other threads.
I'm very tempted to just knock out this particular test on arm*, at
least as a build-time test (it can stay as an autopkgtest, optionally
marked as flaky, if we want that).
smcv
