Tested
heapless RUSTSEC-2026-0574 to 0582: why cargo audit still lists heapless after 0.9.4
RUSTSEC-2026-0574 to RUSTSEC-2026-0582 are nine unsound notices for the heapless crate, and only heapless 0.9.4 is listed as patched. RustSec dated all nine 18 November 2026, the day heapless 0.9.4 was released, and added them to its database on 19 November. None of the nine lists a CVE or GHSA alias. cargo audit 0.23.0 prints them as warnings and exits 0; it exits 1 with --deny unsound or --deny warnings. A direct dependency on heapless 0.9.x clears all nine with one cargo update. A copy of 0.6, 0.7 or 0.8 does not: on 22 November 2026 crates.io had no heapless 0.7.18 or 0.8.1, the heapless repository's v0.7.x and v0.8.x branches still pointed at the commits of 0.7.17 and 0.8.0, and a crate that asks for ^0.7 or ^0.8 cannot be moved to 0.9.4 by cargo update. In the 20 real lockfiles we read, 16 held such a copy, most often because of postcard (7 of the 16) or rstar (4). There are three ways out: update a 0.9.x copy, move or reconfigure the crate that pulls the old copy, or ignore the nine IDs with a scope, a reason and an end date.
Is it a vulnerability in your build? RustSec's own label is unsound, and cargo-audit reports the nine under warnings, with vulnerabilities.found false in its JSON. None of the cargo-audit output we saw says whether your code calls the functions the advisories name, and this page does not either.
Find which crate pulls the old copy
Run these in the workspace the failing job scans. The version in the commands is an example: replace 0.8.0 with the heapless version your Cargo.lock holds, in every command on this page.
grep -A1 '^name = "heapless"$' Cargo.lockThis prints the name and version lines of each heapless entry in the lock. Then ask Cargo who requires the one you care about.
cargo tree --locked --workspace --target all -i heapless@0.8.0The inverted tree puts heapless at the top, then the crates that depend on it, down to one of your own. Three details from our runs:
- With two heapless versions in the lock,
cargo tree -i heaplesswithout a version exits 101 witherror: specification `heapless` is ambiguousand listsheapless@0.7.17andheapless@0.9.3as the forms to use. --workspacematters. On a workspace whose root package does not use heapless and whose member does, the same command without it exited 101 withpackage ID specification `heapless@0.7.17` did not match any packages. With--workspace, the same error had two causes in our runs. Either the version you named is not in theCargo.lockyou ran it against, or it is an optional dependency of one of your own crates that no enabled feature turns on. In a fixture withheapless = { version = "0.8", optional = true },Cargo.lockheld heapless 0.8.0 andcargo auditlisted the nine IDs for it, but the command exited 101 with that error. Adding--all-features, or--featureswith the feature that enables heapless, printed the tree.-e featuresshows the feature path as well, for exampleheapless feature "cas"withpostcard feature "heapless-cas"under it.
cargo tree --locked --workspace --target all -e features -i heapless@0.8.0cargo-deny names the dependent too, when it reports one of the nine: it prints 1 dependent has version requirements that preclude updating:, then the crate and version, then its requirement, for example postcard = 1.2.0 and - heapless = '^0.7.0'.
Who pulls it in, in 20 real projects. On 22 November 2026 we read the root Cargo.lock of 20 projects, taken from a saved list of GitHub pull requests with heapless in the title. We took the first pull request of each owner, in list order, and read the default branch at the last commit before that pull request was opened. We skipped four owners whose repository had no root lockfile or whose lockfile had no heapless. These are projects whose maintainers reacted to the notices, not a sample of everyone who depends on heapless.
- 16 of the 20 held a heapless below 0.9: 2 held 0.6.1, 9 held 0.7.17, 12 held 0.8.0, and several held more than one. Ten held a 0.9.x below 0.9.4, 7 of them next to an old copy. Only 1 was already on 0.9.4.
- The crates that required an old copy, with the number of projects that had them: postcard 7, rstar 4, embassy-sync 2, loro-delta 2, usb-device 2, and a crate of the project's own workspace 2. aimdb-mountain-mqtt, faster-hex and picoserve appeared once each. One of the four rstar projects is the rstar repository's own lockfile.
The nine IDs, and what 0.7.17 and 0.8.0 contain
Each advisory names functions, and most give a version range per function. The [versions] section of each advisory says only patched = [">= 0.9.4"], with no unaffected range. Below is what the heapless source contains for each function the advisories name, in 0.7.17 (the last 0.7 release) and 0.8.0 (the only 0.8 release). We read the crate files downloaded from crates.io, and compiled a one-line probe for each function against each version on x86_64-unknown-linux-gnu with default features. Unless a row says otherwise, the advisory lists the function for versions below 0.9.4.
| ID (RUSTSEC-2026-) | Function the advisory names | In 0.7.17 | In 0.8.0 |
|---|---|---|---|
| 0574 | sorted_ | yes, type is named FindMut | yes, type is named FindMut |
| 0575 | spsc::Spsc::peek | no type Spsc (a) | no type Spsc (a) |
| 0576 | index_ | no | no |
| 0577 | deque::Deque::clear, listed for < 0.9.3 | yes | yes |
| 0577 | history_, listed for < 0.9.3 | yes, type is named HistoryBuffer | yes, type is named HistoryBuffer |
| 0577 | index_, listed for < 0.9.3 | yes | yes |
| 0577 | index_ | no | yes |
| 0577 | index_ | no | no |
| 0577 | vec::IntoIter::drop | yes | yes |
| 0577 | history_ | yes, type is named HistoryBuffer | yes, type is named HistoryBuffer |
| 0578 | sorted_ | no, there is no zeroize feature | no, there is no zeroize feature |
| 0579 | Deque::try_from, listed for >= 0.9.2, < 0.9.4 | no | no |
| 0580 | pool, listed for < 0.8.0 | yes (b) | in the source, outside that range (b) |
| 0580 | arc_pool | in the source (b) | in the source (b) |
| 0580 | box_pool | no | in the source (b) |
| 0580 | object_pool | no | in the source (b) |
| 0581 | deque::Deque::make_ | no | no |
| 0582 | index_ | yes | yes |
How to read the table:
- "yes" means a method of that name exists on that type. In 0.7.17 and 0.8.0 the types are exported at the crate root (
heapless::IndexMap,heapless::Deque,heapless::HistoryBuffer), and the modulesindex_mapandhistory_bufthat the advisories write do not exist there, whiledequeandvecare private. A probe that named each of those modules failed to compile in both versions. - (a) No type named
Spscis in any heapless we downloaded: 0.7.17, 0.8.0, 0.9.1, 0.9.3 or 0.9.4.peekexists onspsc::Queueand onspsc::Consumerin 0.7.17, 0.8.0 and 0.9.4, and the upstream fix for RUSTSEC-2026-0575 editspeekin the queue code. - (b) The pool code is compiled only on some targets. In 0.7.17,
heapless::pool::Poolcompiled on x86_64-unknown-linux-gnu with default features, and thearc_pool!macro needs an ARM cfg or thex86-sync-poolfeature. In 0.8.0 thepoolmodule is compiled only for ARM targets with thearm_llsccfg or 32-bit x86. On x86_64-unknown-linux-gnu with default features,heapless::pooland the three pool macros did not compile in 0.8.0. Other targets were not compiled. - The advisory files carry a
# TODOcomment on the function line of RUSTSEC-2026-0578, RUSTSEC-2026-0579 and RUSTSEC-2026-0582. The 0.9.4 changelog listsTryFromforDequeunder v0.9.2, which matches the>= 0.9.2in RUSTSEC-2026-0579.
Four IDs (0576, 0578, 0579 and 0581) name only functions that are in neither 0.7.17 nor 0.8.0, and RUSTSEC-2026-0579 itself lists its function only from 0.9.2. The scanners list all nine anyway. cargo-audit 0.23.0 listed all nine IDs for 0.6.1, 0.7.17, 0.8.0 and 0.9.3 alike, and an api.osv.dev query for heapless 0.8.0 returned all nine, while one for 0.9.4 returned none. That matches the [versions] section of every ID, which marks every version below 0.9.4 as unpatched. The table says what the crate contains. It does not say whether your build calls any of these functions, or whether a call is dangerous.
A lock with heapless 0.7.17 lists one more notice. The default cas feature brings in atomic-polyfill 1.0.3, which RUSTSEC-2023-0089 marks unmaintained, so cargo audit --deny warnings counted 10 denied warnings for 0.7.17 and still exited 1 after the nine IDs were ignored.
Way out A: update the 0.9.x copy
If your own Cargo.toml asks for heapless 0.9 and the lock holds 0.9.3 (the version we ran this on), one command fixes that copy.
cargo update -p heapless --precise 0.9.4When the lock holds two heapless versions, the bare name is ambiguous. In our run the command above exited 101 with error: specification `heapless` is ambiguous. Name the copy.
cargo update -p heapless@0.9.3 --precise 0.9.4That moved only the 0.9.3 copy: Updating heapless v0.9.3 -> v0.9.4. The 0.7.17 copy from postcard stayed, and cargo-audit still listed the nine IDs for it. cargo update -p heapless@0.7.17 --precise 0.9.4 exited 101 with failed to select a version for the requirement `heapless = "^0.7.0"`, required by package `postcard v1.2.0`.
If your Cargo.toml pins heapless = "=0.9.3", the update exited 101 with failed to select a version for the requirement `heapless = "=0.9.3"`. Change the requirement, to "0.9" for example, and run the update again.
Rust version. heapless 0.9.4 declares rust-version = "1.88" in its Cargo.toml. 0.9.3 and 0.9.2 declare 1.87, and 0.9.1 and 0.9.0 declare none. The changelog says: "Bump MSRV to 1.88 (requires by trybuild dev dependency)". With Cargo 1.87.0, a project on heapless 0.9.3 built. After cargo update -p heapless updated it to 0.9.4, cargo check exited 101 with error: rustc 1.87.0 is not supported by the following packages: heapless@0.9.4 requires rustc 1.88. A manifest that declares rust-version = "1.87" and uses edition 2024 behaved differently in Cargo 1.99.1: plain cargo update -p heapless left 0.9.3 (Locking 0 packages to highest Rust 1.87 compatible versions), and --precise 0.9.4 moved it, printing (requires Rust 1.88). With edition 2021 and the same rust-version, plain cargo update -p heapless moved to 0.9.4.
Way out B: move or reconfigure the crate that pulls the old copy
What we found for the crates that appeared in two or more of the 20 lockfiles. Release dates are from the crates.io API and upstream states are from GitHub, read on 22 November 2026.
postcard (7 projects). postcard 1.2.0, published 21 November 2026 and declaring rust-version = "1.88", adds a heapless-v0_9 feature with postcard::heapless_, to_vec_cobs and to_vec_crc32 as replacements. Its default feature heapless-cas still depends on heapless ^0.7.0. In its docs, 1.2.0 says that heapless 0.7 "is affected by RUSTSEC-2026-0575 and will not be fixed". What we tested on fixtures:
default-features = falseremoved heapless from the lockfile, for 1.2.0 and for 1.1.3.postcard::to_slicecompiled;postcard::to_allocveccompiled with theallocoruse-stdfeature;postcard::to_vecdid not compile. In postcard 1.2.0's source,to_vecandto_vec_cobsare behind theheaplessfeature, as areto_vec_crc32(withuse-crc) and theHVecflavor. We compiled onlyto_vec; other gated items were not tested.- 1.2.0 with
default-features = false, features = ["heapless-v0_9"]left heapless 0.9.4 as the only copy, andpostcard::heapless_compiled.v0_ 9::to_ vec - 1.2.0 with its default features plus
heapless-v0_9left both 0.7.17 and 0.9.4 in the lock, and cargo-audit still listed the nine IDs. - 1.2.0 also has a
heapless-v0_8feature. Withdefault-features = false, features = ["heapless-v0_8"]the lock held heapless 0.8.0, which has no patched release either. - 1.1.3 has no 0.9 option.
rstar (4 projects). rstar 0.14.0, published 1 November 2026 with rust-version = "1.90", requires heapless ^0.9; a fixture with it resolved heapless 0.9.4 and cargo-audit listed nothing. rstar 0.13.x requires ^0.8, declares rust-version = "1.85", and heapless is not optional: default-features = false kept 0.8.0 in the lock. No release or switch for 0.13 at read time. Upstream, a pull request "Bump heapless to 0.9 on the 0.13 maintenance branch" and an issue "Release 0.12.x on heapless 0.9.4 (RUSTSEC-2026-0574 to 0582)" were both open.
embassy-sync (2 projects). embassy-sync 0.8.0, published 20 March 2026, requires heapless ^0.9. 0.6.1 to 0.7.2 require ^0.8 and heapless is not optional: default-features = false on 0.7.2 kept 0.8.0.
loro-delta (2 projects). 1.18.0, published 16 November 2026, still requires heapless ^0.8.0, not optional; default-features = false kept 0.8.0. No release or switch found at read time, and a search of the loro repository for "heapless" found no issue or pull request about heapless 0.9.
usb-device (2 projects). 0.3.2, published 6 March 2024, requires ^0.8, not optional; default-features = false kept 0.8.0. Upstream, the issue "Release 0.3.3 with heapless bumped to 0.9" was open.
A crate of your own workspace (2 projects, both holding 0.8.0). Way out A does not reach this copy: as with postcard's ^0.7.0 above, cargo update cannot move a copy to 0.9.4 when the requirement that pulls it in does not allow 0.9. Leaving 0.8 means changing your own requirement to "0.9" and taking the API changes that heapless lists in its changelog under 0.9.0 and 0.9.1 (0.9.0 itself is yanked); we did not test those code changes. In a two-member fixture with no code, one member on 0.9.3 and the other changed from "0.8" to "0.9", cargo update -p heapless@0.8.0 printed Removing heapless v0.8.0 and left the 0.9.3 copy, which cargo update -p heapless@0.9.3 --precise 0.9.4 then moved.
faster-hex, which appeared once, published 2.0.0 on 20 November 2026 with heapless ^0.9.4 as an optional dependency. Other versions of these crates, and other parents, were not tested. These lists are as of 22 November 2026; check each crate's versions before you rely on them.
Way out C: ignore the nine IDs, with a scope, a reason and an end
cargo-deny can limit an ignore to the crate that depends on heapless (allow) and can end it on a date (expiry). Its changelog lists both under 0.21.0; we ran the block on 0.22.0 and 0.21.0, with the same results. cargo-audit can do neither. Keep your own unsound setting in cargo-deny. At its default, a copy that comes only through a parent already passes (see the table under "What each tool does with the nine"), so you need these entries when you set unsound = "all" or "transitive". The allow value is an example: replace rstar@0.13 with the crate that your cargo tree output shows, and the date with your own.
[advisories]
ignore = [
{ id = "RUSTSEC-2026-0574", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
{ id = "RUSTSEC-2026-0575", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
{ id = "RUSTSEC-2026-0576", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
{ id = "RUSTSEC-2026-0577", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
{ id = "RUSTSEC-2026-0578", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
{ id = "RUSTSEC-2026-0579", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
{ id = "RUSTSEC-2026-0580", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
{ id = "RUSTSEC-2026-0581", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
{ id = "RUSTSEC-2026-0582", reason = "heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9", allow = ["rstar@0.13"], expiry = "2027-02-28" },
]To make a stale entry fail the job, add unused-ignored-advisory = "deny" under [advisories]. Its default in cargo-deny 0.22.0 is warn. Only add it when the entries match something now: at the default unsound scope, with heapless 0.8.0 coming through rstar 0.13.1, the nine entries matched nothing and the setting turned that into nine errors.
What allow matched, with a lock that held the parent at the version shown: rstar, rstar@0.13 and rstar@0.13.1 matched rstar 0.13.1, and rstar@0.14.0 did not. postcard, postcard@1 and postcard@1.2.0 matched postcard 1.2.0, and postcard@1.1.3 did not.
For cargo-audit, add the IDs to .cargo/audit.toml. The example configuration that ships with cargo-audit shows ignore as a list of IDs, so the reason goes in a comment.
[advisories]
# heapless 0.8 reached only through rstar 0.13; remove when rstar moves to heapless 0.9
ignore = [
"RUSTSEC-2026-0574",
"RUSTSEC-2026-0575",
"RUSTSEC-2026-0576",
"RUSTSEC-2026-0577",
"RUSTSEC-2026-0578",
"RUSTSEC-2026-0579",
"RUSTSEC-2026-0580",
"RUSTSEC-2026-0581",
"RUSTSEC-2026-0582",
]cargo audit reads .cargo/audit.toml from the directory you run it in. When we ran it from another directory with --file pointing at a lockfile, the ignore in the config next to that lockfile was not applied: with --deny warnings the run exited 1. For one run, or for the audit action, pass each ID with --ignore. The action passes every ID in its ignore input to cargo audit as --ignore, and describes the input as a comma-separated list. It splits the input on commas only, so keep the commas: given one value holding two IDs on separate lines, cargo audit stopped with malformed advisory ID.
- uses: rustsec/audit-check@v2.0.0
with:
token: ${{ secrets.GITHUB_TOKEN }}
ignore: RUSTSEC-2026-0574,RUSTSEC-2026-0575,RUSTSEC-2026-0576,RUSTSEC-2026-0577,RUSTSEC-2026-0578,RUSTSEC-2026-0579,RUSTSEC-2026-0580,RUSTSEC-2026-0581,RUSTSEC-2026-0582We ran the same arguments the action builds, cargo audit --ignore <ID> ... --json --file ./Cargo.lock, and the JSON listed no warnings for heapless 0.8.0. We did not run the action. Issues it already opened stay open, and it does not open a second one while an issue or pull request with the ID in its title exists.
What each tool prints when the ignore stops matching. We ran both tools with the nine entries in place, first with rstar 0.13.1 (heapless 0.8.0) and then with rstar 0.14.0 (heapless 0.9.4 only). The cargo-deny rows used the block above with unsound = "all" added. The third row also added unused-ignored-advisory = "deny", the fourth changed allow, and the fifth changed the date.
| State | cargo-deny 0.22.0 | cargo-audit 0.23.0 |
|---|---|---|
| heapless 0.8.0 through rstar 0.13.1, entries in place | Exit 0, advisories ok | Exit 0, no warning. Exit 0 with --deny warnings too |
| rstar 0.14.0, heapless 0.9.4 only, entries still in place | Exit 0, and nine warning[advisory-not-detected]: advisory was not encountered | Exit 0, no warning |
Same, with unused-ignored-advisory = "deny" | Exit 1, nine error[advisory-not-detected] | No such setting found in the example configuration |
allow names rstar@0.14.0, the lock has rstar 0.13.1 | Exit 1, nine error[unsound], nine advisory-ignore-disallowed-dependent warnings and nine advisory-ignore-allowed-dependent-missing warnings | Not applicable |
expiry date already past | Exit 1, the nine error[unsound] come back | Not applicable |
Two cautions from the same runs.
- Both tools match the ID, not the version. In a fixture with the project's own heapless 0.9.3 and rstar 0.13.1's 0.8.0, ignoring the nine IDs silenced both copies in cargo-audit and in cargo-deny with
unsound = "all". With theallow = ["rstar@0.13"]entries above, cargo-deny reported the project's own 0.9.3 again, nine errors. Update your own copy first (Way out A), then ignore the parent's. - cargo-audit's
informational_warningssetting is not scoped to heapless. Withinformational_warnings = ["unmaintained"], the value in the example configuration, or with[], our three fixtures with an old heapless reported nounsoundwarning for any crate;--deny unsoundstill exited 1. It hides every unsound notice, not these nine.
What each tool does with the nine
Our guide to RUSTSEC-2026-0543, chrono is unmaintained covers how the same tools treat unmaintained notices; this table is only what differs for unsound. Each cell is one run.
| Tool and setting | heapless 0.8.0 is your direct dependency, one copy | heapless 0.8.0 only through rstar 0.13.1 | Your own 0.9.3 plus 0.8.0 through rstar 0.13.1 |
|---|---|---|---|
cargo audit 0.23.0 | Exit 0. Nine warning blocks, then warning: 9 allowed warnings found | Same | Exit 0. warning: 18 allowed warnings found, both copies |
cargo audit --deny unsound, or --deny warnings | Exit 1. error: 9 denied warnings found! | Same | Exit 1. error: 18 denied warnings found! |
cargo deny check advisories 0.22.0, no deny.toml or [advisories] left at its defaults | Exit 1. Nine error[unsound] | Exit 0. advisories ok | Exit 0. advisories ok, for your own 0.9.3 too |
Same, unsound = "all" | Exit 1 | Exit 1 | Exit 1, 18 errors |
Same, unsound = "workspace" (the default) | Exit 1 | Exit 0 | Exit 0 |
Same, unsound = "transitive" | Exit 0 | Exit 1 | Exit 1, nine errors, for the 0.8.0 copy only |
Same, unsound = "none" | Exit 0 | Exit 0 | Exit 0 |
cargo deny check advisories 0.21.0 at its defaults | Exit 1. Nine error[unsound] | Exit 0 | Exit 0 |
- cargo-deny 0.21.0 and 0.22.0 gave the same
unsoundresults at their defaults in every fixture we ran. With heapless 0.7.17 through postcard, 0.21.0 also exited 1, because of the atomic-polyfill unmaintained notice, which 0.22.0 does not report for a transitive copy. cargo-deny's own table of scopes is in its advisories configuration:unsounddefaults toworkspacein both versions. - The two-copy column held in all five arrangements we ran: heapless declared before or after the parent, the parent being postcard, embassy-sync 0.7.2 or rstar 0.13.1, and
=0.9.3written in the manifest. When one version is both your direct dependency and a parent's, as heapless 0.8.0 with rstar 0.13.1,workspacedid flag it. - The unreported copy was not always a 0.9.3. With your own heapless 0.8.0 next to postcard 1.2.0's 0.7.17, the default also exited 0, and your own 0.8.0 went unreported. With your own 0.7.17 next to rstar 0.13.1's 0.8.0 it went the other way: the default reported your own copy, and
unsound = "transitive"exited 0 without reporting rstar's 0.8.0.unsound = "all"reported every copy in each case. 0.21.0 reported the sameunsounderrors. --deny unmaintaineddid not fail on a lock that held only heapless 0.8.0 (exit 0).[output] deny = ["unsound"]in.cargo/audit.tomlmakescargo auditexit 1, like the flag, but only withquiet = falsebeside it: withdenyalone, 0.23.0 stopped with a parse error that said thequietfield was missing.- The audit action, version 2.0.0, which we read and did not run:
src/countsreporter.ts unsoundapart from vulnerabilities, so the check run it publishes concludessuccessfor them, and on a scheduled run it creates one issue per ID, titled with the ID, a colon and the advisory title, such asRUSTSEC-2026-0580: The treiber stack implementation in heapless had a race condition. cargo deny checkwith no configuration runs every check. In our run it exited 5 and printedadvisories FAILED, bans ok, licenses FAILED, sources ok; runcargo deny check advisoriesto see the nine on their own.
What we ran
We ran the tools in the official Rust Docker image for Linux, pinned by digest, with Cargo 1.99.1, cargo-audit 0.23.0 and cargo-deny 0.22.0; the cargo-deny 0.21.0 rows used that release. Every scanner read the same copy of the RustSec database at commit 55e1c3dfbd83 from 21 November 2026, which includes the 19 November commit 7133866a6b that assigned the nine IDs. The fixtures were small crates with no code, resolved on 22 November 2026: one each on heapless 0.6.1, 0.7.17, 0.8.0, 0.9.3 and 0.9.4, one with heapless 0.9.3 and postcard 1.2.0, and one each on postcard 1.2.0, rstar 0.13.1 and 0.14.0, embassy-sync 0.7.2 and 0.8.0, usb-device 0.3.2, and loro-delta 1.18.0, plus the switch variants named above. Later runs added a crate whose heapless is an optional dependency that no feature enables, and two-copy fixtures with your own heapless 0.8.0 or 0.7.17 next to a parent's copy. The Cargo 1.87.0 result came from a second run with that toolchain. Each result above is one run with one exit code. We did not run Windows, macOS, shells other than bash, targets other than x86_64-unknown-linux-gnu, cargo-audit before 0.23.0, the audit action itself, osv-scanner, or any version of the parent crates other than those named.
Sources
- RUSTSEC-2026-0574 to 0582 on rustsec.org: 0574, 0575, 0576, 0577, 0578, 0579, 0580, 0581 and 0582, read 22 November 2026 at 01:48 UTC.
- RustSec advisory database at commit 55e1c3dfbd83, the advisory files for the nine IDs, read 22 November 2026 at 01:28 UTC; the database head was unchanged at 01:59 UTC. The commit that assigned the IDs is dated 19 November 2026.
- heapless on crates.io: versions, dates and checksums from the crates.io index and API, read 22 November 2026 at 01:29 UTC; the index was unchanged at 01:59 UTC. Crate files for 0.7.17, 0.8.0, 0.9.1, 0.9.3 and 0.9.4 from
static.crates.io, 22 November 2026 at 01:29 UTC. - heapless 0.9.4 Cargo.toml and 0.9.4 changelog on docs.rs, read 22 November 2026 at 01:51 UTC.
- heapless repository: branches
v0.7.xandv0.8.x, read withgit ls-remoteon 22 November 2026 at 01:41 UTC and again at 01:59 UTC. - Parent crates on crates.io: postcard, rstar, embassy-sync, loro-delta, usb-device and faster-hex: index files re-read and unchanged on 22 November 2026 between 01:59 and 02:00 UTC. postcard 1.2.0 README on docs.rs, read 22 November 2026 at 01:51 UTC.
- Upstream states, read 22 November 2026 at 01:40 UTC and again at 01:59 UTC: rstar pull request 274, rstar issue 273 and usb-device issue 175.
- rustsec/audit-check v2.0.0: src/reporter.ts, src/main.ts, src/input.ts and action.yml, read 22 November 2026 at 01:32 UTC. v2.0.0 was the newest tag at that time.
- cargo-deny 0.22.0 advisories configuration and changelog, and the 0.21.0 configuration, read 22 November 2026 at 02:00 UTC.
- cargo-audit 0.23.0 example configuration, read 22 November 2026 at 02:00 UTC.
- OSV: a query to the OSV API for the crates.io package heapless at 0.8.0 and at 0.9.4, 22 November 2026 at 01:48 UTC.
Corrections can be sent to hello@cybernative.ai. We publish dated corrections.