Debian Patches

Status for rust-governor/0.10.4-2

Patch Description Author Forwarded Bugs Origin Last update
auto/auto-remove-features.patch This patch was autogenerated using debcargo. Do not edit manually! no
0001_relax-dependencies.patch Relax dependencies The upstream set the dependencies too strict. Relax some dependencies
because Debian has newer version packaged.
===================================================================		 Ying-Chun Liu (PaulLiu) <paulliu@debian.org> not-needed 2026-10-16
0002-hashmap-use-std-feature.patch Use hashbrown HashMap when std feature is not enabled Governor uses the "no_std" feature to switch its keyed state store
between std::collections::HashMap and hashbrown::HashMap. Debian
removes the "no_std" feature (remove_features in debcargo.toml), so a
build without the "std" feature (e.g. with only the dashmap/jitter/
quanta features, which is a no_std build) would always select
std::collections::HashMap and fail to compile with
"cannot find module or crate std".
Gate the HashMap selection on the "std" feature instead, which mirrors
the feature gating used everywhere else in governor's source.
===================================================================		 Ying-Chun Liu (PaulLiu) <paulliu@debian.org> not-needed 2026-10-20
0003-drop-bench.patch Drop the criterion benchmark The criterion benchmark target (governor_criterion_benches) only
compiles when all of the std, dashmap and quanta features are enabled
at once, so it breaks autopkgtest runs where each feature is tested on
its own (e.g. --no-default-features --features std fails with
"cannot find QuantaClock/DashMapStateStore in state::keyed"). It also
pulls in a large dev-dependency tree. Drop the benchmark target.
===================================================================		 Ying-Chun Liu (PaulLiu) <paulliu@debian.org> not-needed 2026-10-20
0004-jitter-test-use-alloc-format.patch Use alloc::format in the jitter test module The jitter test module is gated on `not(feature = "no_std")`. Since
Debian removes the "no_std" feature (see remove_features in
debcargo.toml), this condition is always true and the module is
compiled even for no_std builds (e.g. `--no-default-features
--features jitter`). In a no_std crate the `format!` macro is not in
scope, so the `uniform_sampler_coverage` test fails to compile with
"cannot find macro `format` in this scope". lib.rs declares
`extern crate alloc;` unconditionally, so import `alloc::format` in
the test module to make it build in both std and no_std configurations.
===================================================================		 Ying-Chun Liu (PaulLiu) <paulliu@debian.org> not-needed 2026-10-20
0005-future-test-relax-timing-threshold.patch Relax MAX_TEST_RUN_DURATION in the future tests The `proceeds*` tests in tests/future.rs assert that
`until_ready`/`until_n_ready`/`until_key_ready`/`until_key_n_ready`
return within MAX_TEST_RUN_DURATION when the limiter has spare
capacity. The threshold of 200 microseconds is far too tight: under
CI/autopkgtest load the tests sporadically measure several hundred
microseconds of scheduling noise (e.g. the `proceeds_keyed_n` test
measured 664us) and fail spuriously. A genuine blocking wait is
asserted to take at least 100ms by the `pauses*` tests, so bump the
threshold to 10ms; this still reliably distinguishes the "proceeds"
case from the "pauses" case while being robust to scheduler noise.
===================================================================		 Ying-Chun Liu (PaulLiu) <paulliu@debian.org> not-needed 2026-10-20
0006-memory-leaks-fix-ru-maxrss-type.patch Cast libc's ru_maxrss to i64 in the memory-leaks test The resident_memory_size() helper in tests/memory_leaks.rs declares
an i64 return type but returns out.ru_maxrss directly. The width of
that field follows the target's `long` type: it is i64 on 64-bit
targets (amd64, arm64, ...) but i32 on 32-bit targets such as i386,
where the test fails to compile with error[E0308]: mismatched types.
Cast the field to i64 so the test builds on all architectures; the
value is a non-negative count of kilobytes, so the widening
conversion is lossless.
===================================================================		 Ying-Chun Liu (PaulLiu) <paulliu@debian.org> invalid 2026-10-29
0007-future-test-relax-pauses-threshold.patch Relax MIN_TEST_RUN_DURATION in the future pauses tests The `pauses*` tests in tests/future.rs assert that a blocking wait
(until_ready/until_n_ready/until_key_ready/until_key_n_ready)
actually takes at least 100ms when the limiter is exhausted. The
wait is scheduled to land right on that 100ms deadline, so under
CI/autopkgtest load the wall-clock measurement can occasionally come
in a few tens of microseconds short (e.g. the pauses_keyed_n test
measured 99.936ms and failed spuriously). Relax the lower bound to
50ms: this still unambiguously distinguishes the blocking "pauses"
case from the non-blocking "proceeds" case (whose upper bound,
MAX_TEST_RUN_DURATION, is 10ms after 0005-future-test-relax-timing-
threshold.patch) while being robust to timer and scheduler noise.
===================================================================		 Ying-Chun Liu (PaulLiu) <paulliu@debian.org> not-needed 2026-10-29

All known versions for source package 'rust-governor'

Links