Quelle  critical-bugs.md   Sprache: unbekannt

Compilation of critical bugs in stable releases of Coq
======================================================

This file recollects knowledge about critical bugs found in Coq since version 8.0.

<!-- generate using dev/tools/markdown-toc -->

## Table of Contents

  - [Non fixed bugs](#non-fixed-bugs)
      - [buffer overflow on large records and closures (infinite loop with OCaml 5)](#buffer-overflow-on-large-records-and-closures-infinite-loop-with-ocaml-5)
      - [memory corruption by evaluating on ill-typed terms (obtained from unsafe tactics)](#memory-corruption-by-evaluating-on-ill-typed-terms-obtained-from-unsafe-tactics)
      - [kernel and checker accept incorrect name aliasing information](#kernel-and-checker-accept-incorrect-name-aliasing-information)
      - [coqchk checks too little about primitive declarations](#coqchk-checks-too-little-about-primitive-declarations)
      - [Print Assumptions + Parameter Inline fails to report some inconsistent flags](#print-assumptions-parameter-inline-fails-to-report-some-inconsistent-flags)
      - [Print Assumptions does not report Unset Universe Checking used during functor application](#print-assumptions-does-not-report-unset-universe-checking-used-during-functor-application)
  - [Fixed bugs](#fixed-bugs)
    - [Typing constructions](#typing-constructions)
      - [substitution missing in the body of a let](#substitution-missing-in-the-body-of-a-let)
      - [missing lift in checking guard](#missing-lift-in-checking-guard)
      - [de Bruijn indice bug in checking guard of nested cofixpoints](#de-bruijn-indice-bug-in-checking-guard-of-nested-cofixpoints)
      - [de Bruijn indice bug in computing allowed elimination principle](#de-bruijn-indice-bug-in-computing-allowed-elimination-principle)
      - [bug in Prop<=Set conversion which made Set identifiable with Prop, preventing a proof-irrelevant interpretation of Prop](#bug-in-propset-conversion-which-made-set-identifiable-with-prop-preventing-a-proof-irrelevant-interpretation-of-prop)
      - [incorrect abstraction of sort variables in relevance marks on opaque constants](#incorrect-abstraction-of-sort-variables-in-relevance-marks-on-opaque-constants)
      - [guard checker forgot to check non-structural arguments of fixpoint](#guard-checker-forgot-to-check-non-structural-arguments-of-fixpoint)
      - [guard checker incorrectly detects match on match as returning a subterm](#guard-checker-incorrectly-detects-match-on-match-as-returning-a-subterm)
      - [guard checker does incorrect reduction across inner fixpoint, accepting wrong fixpoints](#guard-checker-does-incorrect-reduction-across-inner-fixpoint-accepting-wrong-fixpoints)
    - [Module system](#module-system)
      - [missing universe constraints in typing "with" clause of a module type](#missing-universe-constraints-in-typing-with-clause-of-a-module-type)
      - [universe constraints for module subtyping not stored in vo files](#universe-constraints-for-module-subtyping-not-stored-in-vo-files)
      - [module subtyping disrespected squashing status of inductives](#module-subtyping-disrespected-squashing-status-of-inductives)
      - [Functor inlining drops universe substitution](#functor-inlining-drops-universe-substitution)
      - [Primitives are incorrectly considered convertible to anything by module subtyping](#primitives-are-incorrectly-considered-convertible-to-anything-by-module-subtyping)
      - [missing substitution when strengthening functors](#missing-substitution-when-strengthening-functors)
    - [Universes](#universes)
      - [issue with two parameters in the same universe level](#issue-with-two-parameters-in-the-same-universe-level)
      - [universe polymorphism can capture global universes](#universe-polymorphism-can-capture-global-universes)
      - [template polymorphism not collecting side constraints on the universe level of a parameter](#template-polymorphism-not-collecting-side-constraints-on-the-universe-level-of-a-parameter)
      - [more template polymorphism missing constraints](#more-template-polymorphism-missing-constraints)
      - [universe constraints erroneously discarded when forcing an asynchronous proof containing delayed monomorphic constraints inside a universe polymorphic section](#universe-constraints-erroneously-discarded-when-forcing-an-asynchronous-proof-containing-delayed-monomorphic-constraints-inside-a-universe-polymorphic-section)
      - [Set+2 incorrectly simplified to Set+1](#set2-incorrectly-simplified-to-set1)
      - [variance inference for section universes ignored use of section universes in inductives and axioms defined before the inductive being inferred](#variance-inference-for-section-universes-ignored-use-of-section-universes-in-inductives-and-axioms-defined-before-the-inductive-being-inferred)
    - [Primitive projections](#primitive-projections)
      - [check of guardedness of extra arguments of primitive projections missing](#check-of-guardedness-of-extra-arguments-of-primitive-projections-missing)
      - [records based on primitive projections became possibly recursive without the guard condition being updated](#records-based-on-primitive-projections-became-possibly-recursive-without-the-guard-condition-being-updated)
      - [incorrect checking of subtyping with algebraic universes](#incorrect-checking-of-subtyping-with-algebraic-universes)
    - [Conversion machines](#conversion-machines)
      - [the invariant justifying some optimization was wrong for some combination of sharing side effects](#the-invariant-justifying-some-optimization-was-wrong-for-some-combination-of-sharing-side-effects)
      - [collision between constructors when more than 256 constructors in a type](#collision-between-constructors-when-more-than-256-constructors-in-a-type)
      - [wrong universe constraints](#wrong-universe-constraints)
      - [missing pops in executing 31bit arithmetic](#missing-pops-in-executing-31bit-arithmetic)
      - [primitive integer emulation layer on 32 bits not robust to garbage collection](#primitive-integer-emulation-layer-on-32-bits-not-robust-to-garbage-collection)
      - [broken long multiplication primitive integer emulation layer on 32 bits](#broken-long-multiplication-primitive-integer-emulation-layer-on-32-bits)
      - [broken addmuldiv operation for large shifts](#broken-addmuldiv-operation-for-large-shifts)
      - [translation of identifier from Coq to OCaml was not bijective, leading to identify True and False](#translation-of-identifier-from-coq-to-ocaml-was-not-bijective-leading-to-identify-true-and-false)
      - [stuck primitive projections computed incorrectly by native_compute](#stuck-primitive-projections-computed-incorrectly-by-native_compute)
      - [incorrect De Bruijn handling when inferring the relevance mark for a lambda](#incorrect-de-bruijn-handling-when-inferring-the-relevance-mark-for-a-lambda)
      - [buffer overflow on large accumulators](#buffer-overflow-on-large-accumulators)
      - [buffer overflow, arbitrary code execution on floating-point operations](#buffer-overflow-arbitrary-code-execution-on-floating-point-operations)
      - [arbitrary code execution on irreducible PArray.set](#arbitrary-code-execution-on-irreducible-parrayset)
      - [arbitrary code execution on arrays of floating point numbers](#arbitrary-code-execution-on-arrays-of-floating-point-numbers)
      - [conversion of Prod / Prod values was comparing the wrong components](#conversion-of-prod-prod-values-was-comparing-the-wrong-components)
      - [η-expansion of cofixpoints was performed in the wrong environment](#-expansion-of-cofixpoints-was-performed-in-the-wrong-environment)
      - [conversion would compare the mutated version of primitive arrays instead of undoing mutation where needed](#conversion-would-compare-the-mutated-version-of-primitive-arrays-instead-of-undoing-mutation-where-needed)
      - [tactic code could mutate a global cache of values for section variables](#tactic-code-could-mutate-a-global-cache-of-values-for-section-variables)
      - [incorrect handling of universe polymorphism](#incorrect-handling-of-universe-polymorphism)
    - [Side-effects](#side-effects)
      - [polymorphic side-effects inside monomorphic definitions incorrectly handled as not inlined](#polymorphic-side-effects-inside-monomorphic-definitions-incorrectly-handled-as-not-inlined)
    - [Forgetting unsafe flags](#forgetting-unsafe-flags)
      - [unsafe typing flags used inside a section would not be reported by Print Assumptions after closing the section](#unsafe-typing-flags-used-inside-a-section-would-not-be-reported-by-print-assumptions-after-closing-the-section)
    - [Conflicts with axioms in library](#conflicts-with-axioms-in-library)
      - [axiom of description and decidability of equality on real numbers in library Reals was inconsistent with impredicative Set](#axiom-of-description-and-decidability-of-equality-on-real-numbers-in-library-reals-was-inconsistent-with-impredicative-set)
      - [guard condition was unknown to be inconsistent with propositional extensionality in library Sets](#guard-condition-was-unknown-to-be-inconsistent-with-propositional-extensionality-in-library-sets)
      - [incompatibility axiom of choice and excluded-middle with elimination of large singletons to Set](#incompatibility-axiom-of-choice-and-excluded-middle-with-elimination-of-large-singletons-to-set)
      - [Incorrect specification of PrimFloat.leb](#incorrect-specification-of-primfloatleb)
      - [Incorrect implementation of SFclassify.](#incorrect-implementation-of-sfclassify)
      - [nativenorm reading back closures as arbitrary floating-point values](#nativenorm-reading-back-closures-as-arbitrary-floating-point-values)
      - [guard condition issue made it inconsistent with univalence](#guard-condition-issue-made-it-inconsistent-with-univalence)
    - [Deserialization](#deserialization)
      - [deserialization of .vo data not properly checked](#deserialization-of-vo-data-not-properly-checked)
  - [Probably non exploitable fixed bugs](#probably-non-exploitable-fixed-bugs)
      - [bug in 31bit arithmetic](#bug-in-31bit-arithmetic)

## Non fixed bugs

#### buffer overflow on large records and closures (infinite loop with OCaml 5)

- component: VM reduction machine
- introduced: 8.1
- impacted versions: 8.1-NOW
- impacted coqchk versions: none (no VM in coqchk)
- fixed in: NONE
- found by: Dolan, Roux, Melquiond
- GH issue number: ocaml/ocaml#6385, coq/coq#13439
- exploit: ??
- risk: requires very large number of arguments, fix block size or nested letins

#### memory corruption by evaluating on ill-typed terms (obtained from unsafe tactics)

- component: VM and native reduction machines
- introduced: 8.1
- impacted versions: 8.1-NOW
- impacted coqchk versions: none (no VM or native in coqchk)
- fixed in: NONE
- found by: Gaëtan Gilbert, Andres Erbsen
- GH issue number: coq/coq#16891
- exploit: requires a memory corruption to craft something that doesn't just SIGSEV
- risk: could be activated by chance but unlikely to produce anything
        other than SIGSEV outside a deliberate attack

#### kernel and checker accept incorrect name aliasing information

- component: name handling / typechecker
- introduced: a long time ago
- impacted versions: -NOW
- impacted coqchk versions: same
- fixed in: NONE
- found by: Pierre-Marie Pédrot
- GH issue number: coq/coq#7609
- exploit: see issue (requires a plugin or hand crafted .vo file)
- risk: low

#### coqchk checks too little about primitive declarations

- component: primitive types and operators
- introduced: v8.10 (#6914 primitive integers)
- impacted versions: coqchk only
- impacted coqchk versions: V8.10-NOW
- fixed in: NONE
- found by: Gaëtan Gilbert
- GH issue number: coq/coq#12439
- exploit: not fully worked out, requires crafted .vo file
- risk: none (requires crafted .vo file)

#### Print Assumptions + Parameter Inline fails to report some inconsistent flags

- component: module functors
- introduced: coq/coq#79
- impacted versions: V8.6-NOW
- impacted coqchk versions: none
- found by: Jason Gross
- GH issue number: coq/coq#12155
- exploit: see issue
- risk: moderate if not using coqchk, none if using coqchk (coqchk rejects the produced file)

#### Print Assumptions does not report Unset Universe Checking used during functor application

- component: module functors
- introduced: v8.11 (#10291) or earlier
- impacted versions: V8.11-NOW
- impacted coqchk versions: none
- found by: Gaëtan Gilbert
- GH issue number: coq/coq#16646
- exploit: see issue
- risk: moderate if not using coqchk, none if using coqchk (coqchk rejects the produced file)

## Fixed bugs

### Typing constructions

#### substitution missing in the body of a let

- component: "match"
- introduced: ?
- impacted released versions: V8.3-V8.3pl2, V8.4-V8.4pl4
- impacted development branches: none
- impacted coqchk versions: ?
- fixed in: master/trunk/v8.5 ([e583a79b5](https://github.com/coq/coq/commit/e583a79b5a0298fd08f34305cc876d5117913e95), 22 Nov 2015, Herbelin), v8.4 ([525056f1](https://github.com/coq/coq/commit/525056f1a630426b78668ab583e228c25b492c35), 22 Nov 2015, Herbelin), v8.3 ([4bed0289](https://github.com/coq/coq/commit/4bed0289d66e6e413ccdea7a33dc747c83bce92e), 22 Nov 2015, Herbelin)
- found by: Herbelin
- exploit: test-suite/success/Case22.v
- GH issue number: ?
- risk: ?

#### missing lift in checking guard

- component: fixpoint, guard
- introduced: probably from V5.10
- impacted released versions: probably V5-V7, V8.0-V8.0pl4, V8.1-V8.1pl4
- impacted development branches: v8.0 ?
- impacted coqchk versions: ?
- fixed in: master/trunk/v8.2 ([ff45afa8](https://github.com/coq/coq/commit/ff45afa83a9235cbe33af525b6b0c7985dc7e091), r11646, 2 Dec 2008, Barras), v8.1 ([f8e7f273](https://github.com/coq/coq/commit/f8e7f273f2e6009c3c0f0eee47c33542a6fdf361), r11648, 2 Dec 2008, Barras)
- found by: Barras
- exploit: test-suite/failure/guard.v
- GH issue number: none
- risk: unprobable by chance

#### de Bruijn indice bug in checking guard of nested cofixpoints

- component: cofixpoint, guard
- introduced: after V6.3.1, before V7.0
- impacted released versions: V8.0-V8.0pl4, V8.1-V8.1pl4, V8.2-V8.2pl2, V8.3-V8.3pl2, V8.4-V8.4pl4
- impacted development branches: none
- impacted coqchk versions: ?
- fixed in: master ([9f81e2c36](https://github.com/coq/coq/commit/9f81e2c360c2be764e71d21ed7c266ee6e8a88c5), 10 Apr 2014, Dénès), v8.4 ([f50ec9e7d](https://github.com/coq/coq/commit/f50ec9e7dbd082c9a465aedda25427d93e12cabe), 11 Apr 2014, Dénès), v8.3 ([40c0fe7f4](https://github.com/coq/coq/commit/40c0fe7f44b3c99bec5188e01197c8a77348a4ee), 11 Apr 2014, Dénès), v8.2 ([06d66df8c](https://github.com/coq/coq/commit/06d66df8c713307625b1c40c054ca06c00ff74b3), 11 Apr 2014, Dénès), v8.1 ([977afae90](https://github.com/coq/coq/commit/977afae90c4e2aa974232b0c664346db72aadaa3), 11 Apr 2014, Dénès), v8.0 ([f1d632992](https://github.com/coq/coq/commit/f1d632992e33a74b30e79271bd3748d69c5a2152), 29 Nov 2015, Herbelin, backport)
- found by: Dénès
- exploit: ?
- GH issue number: none ?
- risk: ?

#### de Bruijn indice bug in computing allowed elimination principle

- component: inductive types, elimination principle
- introduced: 23 May 2006, [9c2d70b](https://github.com/coq/coq/commit/9c2d70b91341552e964979ba09d5823cc023a31c), r8845, Herbelin (part of template polymorphism)
- impacted released versions: V8.1-V8.1pl4, V8.2-V8.2pl2, V8.3-V8.3pl2, V8.4-V8.4pl4
- impacted development branches: none
- impacted coqchk versions: ?
- fixed in: master ([8a01c3685](https://github.com/coq/coq/commit/8a01c36850353c1875383cbc788cec9c42590b57), 24 Jan 2014, Dénès), v8.4 ([8a01c3685](https://github.com/coq/coq/commit/8a01c36850353c1875383cbc788cec9c42590b57), 25 Feb 2014, Dénès), v8.3 ([2b3cc4f85](https://github.com/coq/coq/commit/2b3cc4f85cc134fe58c21d720851e275e6a77ea0), 25 Feb 2014, Dénès), v8.2 ([459888488](https://github.com/coq/coq/commit/4598884884d6db00c485189e3a3b793b05814928), 25 Feb 2014, Dénès), v8.1 ([79aa20872](https://github.com/coq/coq/commit/79aa208728420747a933f38b3aa101c92f4dcde0), 25 Feb 2014, Dénès)
- found by: Dénès
- exploit: see coq/coq#3211
- GH issue number: coq/coq#3211
- risk: ?

#### bug in Prop<=Set conversion which made Set identifiable with Prop, preventing a proof-irrelevant interpretation of Prop

- component: universe subtyping
- introduced: V8.2 ([bba897d5f](https://github.com/coq/coq/commit/bba897d5fd964bef0aa10102ef41cee1ac5fc3bb), 12 May 2008, Herbelin)
- impacted released versions: V8.2-V8.2pl2
- impacted development branches: none
- impacted coqchk versions: ?
- fixed in: master/trunk ([679801](https://github.com/coq/coq/commit/679801623c1f55d0081f952c2094c3572fa39d4f), r13450, 23 Sep 2010, Glondu), v8.3 ([309a53f2](https://github.com/coq/coq/commit/309a53f2e1aa9b2a39654cf5fa23eb632a04c22f), r13449, 22 Sep 2010, Glondu), v8.2 (41ea5f08, r14263, 6 Jul 2011, Herbelin, backport)
- found by: Georgi Guninski
- exploit: test-suite/failure/prop_set_proof_irrelevance.v
- GH issue number: none?
- risk: ?

#### incorrect abstraction of sort variables in relevance marks on opaque constants

and lack of checking of relevance marks on constants in coqchk

- component: sort polymorphism / proof irrelevance
- introduced: V8.10 for the coqchk bug, V8.19 for the coqc bug
- impacted released versions: V8.19.0
- impacted coqchk: versions: V8.10-V8.19.0
- fixed in: V8.19.1, V8.20
- found by: Gaëtan Gilbert
- exploit / GH issue: [#18629](https://github.com/coq/coq/issues/18629)
- risk: low (requires specific plugin code unlikely to be found in non malicious plugin)

#### guard checker forgot to check non-structural arguments of fixpoint

- component: guard checking
- introduced: V8.16 ([#15434](https://github.com/rocq-prover/rocq/pull/15434))
- impacted released versions: V8.16 to V9.0.0
- impacted coqchk versions: Same
- fixed in: V9.0.1, V9.1.0
- found by: ccz181078
- exploit / GH issue: [#20413](https://github.com/rocq-prover/rocq/issues/20413)
- risk: significant (bad terms may be generated by Program machinery)

#### guard checker incorrectly detects match on match as returning a subterm

- component: guard checking
- introduced: V8.16 ([#15434](https://github.com/rocq-prover/rocq/pull/15434))
- impacted released versions: V8.16 to V9.0.0
- impacted coqchk versions: Same
- fixed in: V9.0.1, V9.1.0
- found by: Yann Leray
- exploit / GH issue: [#20455](https://github.com/rocq-prover/rocq/issues/20455)
- risk: unknown (no development in CI was affected)

#### guard checker does incorrect reduction across inner fixpoint, accepting wrong fixpoints

- component: guard checking
- introduced: V8.16 ([#15434](https://github.com/rocq-prover/rocq/pull/15434))
- impacted released versions: V8.16 to V9.0.0
- impacted coqchk versions: Same
- fixed in: V9.0.1, V9.1.0
- found by: Yann Leray
- exploit / GH issue: [#20555](https://github.com/rocq-prover/rocq/issues/20555)
- risk: unknown (no development in CI was affected)

### Module system

#### missing universe constraints in typing "with" clause of a module type

- component: modules, universes
- introduced: ?
- impacted released versions: V8.3-V8.3pl2, V8.4-V8.4pl6; unclear for V8.2 and previous versions
- impacted development branches: none
- impacted coqchk versions: ?
- fixed in: master/trunk ([d4869e059](https://github.com/coq/coq/commit/d4869e059bfb73d99e1f5ef1b0a1f0906fa27056), 2 Oct 2015, Sozeau), v8.4 ([40350ef3b](https://github.com/coq/coq/commit/40350ef3b34b0be9d5ceddde772218c2f2dafe32), 9 Sep 2015, Sozeau)
- found by: Dénès
- exploit: test-suite/bugs/bug_4294.v
- GH issue number: coq/coq#4294
- risk: ?

#### universe constraints for module subtyping not stored in vo files

- component: modules, universes
- introduced: presumably 8.2 ([b3d3b56](https://github.com/coq/coq/commit/b3d3b566c5b5f34ab518c587f62530abde131be8))
- impacted released versions: 8.2, 8.3, 8.4
- impacted development branches: v8.5
- impacted coqchk versions: none
- fixed in: v8.2 ([c1d9889](https://github.com/coq/coq/commit/c1d988904483eb1f3a8917ea08fced1240e3844b)), v8.3 (8056d02), v8.4 ([a07deb4](https://github.com/coq/coq/commit/a07deb4eac1d5f886159784ef5d8d006892be547)), trunk ([0cd0a3e](https://github.com/coq/coq/commit/0cd0a3ecdc7f942da153c59369ca3572bd18dd10)) Mar 5, 2014, Tassi
- found by: Tassi by running coqchk on the mathematical components library
- exploit: requires multiple files, no test provided
- GH issue number: coq/coq#3243
- risk: could be exploited by mistake

#### module subtyping disrespected squashing status of inductives

- component: modules, universes, inductives
- introduced: probably 7.4 ([1296520](https://github.com/coq/coq/commit/12965209478bd99dfbe57f07d5b525e51b903f22))
- impacted released versions: until 8.15.0
- impacted coqchk versions: none
- fixed in: 8.15.1, 8.16
- found by: Pédrot
- exploit: see GitHub issue
- GH issue number: coq/coq#15838
- risk: unlikely (caught by coqchk, needs Unset Elimination Schemes in the module type)

#### Functor inlining drops universe substitution

- component: Modules
- introduced: ?
- impacted released versions: ??-V8.8.0
- impacted coqchk versions: same? not sure if coqchk has a this bug
- fixed in: V8.8.1, V8.9.0 (#7616)
- found by: Pierre-Marie Pédrot
- GH issue number: coq/coq#7615
- exploit: see issue
- risk: medium

#### Primitives are incorrectly considered convertible to anything by module subtyping

- component: modules, primitive types
- introduced: 8.11
- impacted released versions: V8.11.0-V8.18.0
- impacted coqchk versions: same
- fixed in: V8.19.0
- found by: Gaëtan Gilbert
- GH issue number: coq/coq#18503
- exploit: see issue
- risk: high if there is a Primitive in a Module Type, otherwise low

#### Missing substitution when strengthening functors

- component: modules
- introduced: 8.5 for the kernel (c5b699f), 8.10 for the checker (#8773)
- impacted released versions: 8.5-9.0.0
- impacted coqchk version: 8.10-9.0.0
- fixed in: V9.0.1
- found by: Pierre-Marie Pédrot
- GH issue number: rocq-prover/rocq#21051
- exploit: see issue
- risk: could be exploited by mistake when using heavy module machinery

### Universes

#### issue with two parameters in the same universe level

- component: template polymorphism
- introduced: 23 May 2006, [9c2d70b](https://github.com/coq/coq/commit/9c2d70b91341552e964979ba09d5823cc023a31c), r8845, Herbelin
- impacted released versions: V8.1-V8.1pl4, V8.2-V8.2pl2, V8.3-V8.3pl2
- impacted development branches: none
- impacted coqchk versions: ?
- fixed in: trunk/master/v8.4 ([8082d1faf](https://github.com/coq/coq/commit/8082d1faf85a0ab29f6c144a137791902a4e9c1f), 5 Oct 2011, Herbelin), V8.3pl3 ([bb582bca2](https://github.com/coq/coq/commit/bb582bca2ca3fd94df01aad8d8070f8d129b25b3), 5 Oct 2011, Herbelin), v8.2 branch ([3333e8d3](https://github.com/coq/coq/commit/3333e8d3387b3bc4d3ceb75aad853f8e455af444), 5 Oct 2011, Herbelin), v8.1 branch ([a8fc2027](https://github.com/coq/coq/commit/a8fc2027258e1fb2defd05344e1249374f6e4e19), 5 Oct 2011, Herbelin),
- found by: Barras
- exploit: test-suite/failure/inductive.v
- GH issue number: none
- risk: unlikely to be activated by chance

#### universe polymorphism can capture global universes

- component: universe polymorphism
- impacted released versions: V8.5 to V8.8
- impacted coqchk versions: V8.5 to V8.9
- fixed in: [ec4aa4971f](https://github.com/coq/coq/commit/ec4aa4971f7789eeccec2f38f2bb7ec976f87ede) ([58e1d0f200](https://github.com/coq/coq/commit/58e1d0f2006f3243cbf7b57a9858f5119ffea666) for the checker)
- found by: Gaëtan Gilbert
- exploit: test-suite/misc/poly-capture-global-univs
- GH issue number: coq/coq#8341
- risk: unlikely to be activated by chance (requires a plugin)

#### template polymorphism not collecting side constraints on the universe level of a parameter

this is a general form of the previous issue about template
polymorphism exploiting other ways to generate untracked constraints

- component: template polymorphism
- introduced: morally at the introduction of template polymorphism, 23
- May 2006, [9c2d70b](https://github.com/coq/coq/commit/9c2d70b91341552e964979ba09d5823cc023a31c), r8845, Herbelin
- impacted released versions: at least V8.4-V8.4pl6, V8.5-V8.5pl3,
  V8.6-V8.6pl2, V8.7.0-V8.7.1, V8.8.0-V8.8.1, V8.9.0-V8.9.1, in theory
  also V8.1-V8.1pl4, V8.2-V8.2pl2, V8.3-V8.3pl2 but not exploit found
  there yet (an exploit using a plugin to force sharing of universe
  level is in principle possible though)
- impacted development branches: all from 8.4 to 8.9 at the time of writing and suspectingly also all from 8.1 to 8.4 if a way to create untracked constraints can be found
- impacted coqchk versions: a priori all (tested with V8.4 and V8.9 which accept the exploit)
- fixed in: V8.10.0 ([eb3f8225a2](https://github.com/coq/coq/commit/eb3f8225a286aef3a57ad876584b4a927241ff69), PR coq/coq#9918, Aug 2019, Dénès and Sozeau)
- found by: Gilbert using explicit sharing of universes, exploit found for 8.5-8.9 by Pédrot, other variants generating sharing using sections, or using ltac tricks by Sozeau, exploit in 8.4 by Herbelin and Jason Gross by adding new tricks to Sozeau's variants
- exploit: test-suite/failure/Template.v
- GH issue number: coq/coq#9294
- risk: moderate risk to be activated by chance

#### more template polymorphism missing constraints

using the same universe in the parameters and the constructor
arguments of a template polymorphic inductive (using named universes
in modern Coq, or unification tricks in older Coq) produces implicit
equality constraints not caught by the previous template polymorphism
fix.

- component: template polymorphism
- introduced: same as the previous template polymorphism bug, morally
  from V8.1, first verified impacted version V8.5 (the universe
  unification is sufficiently different in V8.4 to prevent our trick
  from working)
- fixed in: expected in 8.10.2, 8.11+beta, master (#11128, Nov 2019, Gilbert)
- found by: Gilbert
- exploit: test-suite/bugs/bug_11039.v
- GH issue number: coq/coq#11039
- risk: moderate risk (found by investigating coq/coq#10504)

#### universe constraints erroneously discarded when forcing an asynchronous proof containing delayed monomorphic constraints inside a universe polymorphic section

- component: universe polymorphism, asynchronous proofs
- introduced: between 8.4 and 8.5 by merging the asynchronous proofs feature branch and universe polymorphism one
- impacted released versions: V8.5-V8.10
- impacted development branches: none
- impacted coqchk versions: immune
- fixed in: coq/coq#10664
- found by: Pédrot
- exploit: no test
- GH issue number: none
- risk: unlikely to be triggered in interactive mode, not present in batch mode (i.e. coqc)

#### Set+2 incorrectly simplified to Set+1

- component: algebraic universes
- introduced: V8.10 (with the SProp commit [7550876976](https://github.com/coq/coq/commit/75508769762372043387c67a9abe94e8f940e80a))
- impacted released versions: V8.10.0 V8.10.1 V8.10.2
- impacted coqchk versions: same
- fixed in: coq/coq#11422
- found by: Gilbert
- exploit: see PR (custom application of Hurkens to get around the refreshing at elaboration)
- GH issue number: see PR
- risk: unlikely to be triggered through the vernacular (the system "refreshes" algebraic
    universes such that +2 increments do not appear), mild risk from plugins which manipulate
    algebraic universes.

#### variance inference for section universes ignored use of section universes in inductives and axioms defined before the inductive being inferred

- component: cumulative inductives and sections
- introduced: V8.12 ([73c3b87463](https://github.com/coq/coq/commit/73c3b874633d6f6f8af831d4a37d0c1ae52575bc))
- impacted released versions: V8.12 to V8.15 including patch releases
- impacted coqchk versions: none
- fixed in: V8.16 coq/coq#15950 ([118ffbc010](https://github.com/coq/coq/commit/118ffbc010ce53ebd45baa42edd28335301ca9a5))
- found by: Gilbert and Pédrot
- exploit: see coq/coq#15916
- risk: could be used inadvertently in developments with complex universe usage, only when using cumulative inductives declared in sections. coqchk still works.

### Primitive projections

#### check of guardedness of extra arguments of primitive projections missing

- component: primitive projections, guard condition
- introduced: 6 May 2014, [a4043608f](https://github.com/coq/coq/commit/a4043608f704f026de7eb5167a109ca48e00c221), Sozeau
- impacted released versions: V8.5-V8.5pl2,
- impacted development branches: none
- impacted coqchk versions: ?
- fixed in: trunk/master/v8.5 ([ba00867d5](https://github.com/coq/coq/commit/ba00867d515624aee734d998bfbe3880f559d907), 25 Jul 2016, Sozeau)
- found by: Sozeau, by analyzing bug report coq/coq#4876
- exploit: to be done (?)
- GH issue number: coq/coq#4876
- risk: consequence of bug found by chance, unlikely to be exploited by chance (MS?)

#### records based on primitive projections became possibly recursive without the guard condition being updated

- component: primitive projections, guard condition
- introduced: 10 Sep 2014, [6624459e4](https://github.com/coq/coq/commit/6624459e492164b3d189e3518864379ff985bf8c), Sozeau (?)
- impacted released versions: V8.5
- impacted development branches: none
- impacted coqchk versions: ?
- fixed in: trunk/master/v8.5 ([120053a50](https://github.com/coq/coq/commit/120053a50f87bd53398eedc887fa5e979f56f112), 4 Mar 2016, Dénès)
- found by: Dénès exploiting bug coq/coq#4588
- exploit: test-suite/bugs/bug_4588.v
- GH issue number: coq/coq#4588
- risk: ?

#### incorrect checking of subtyping with algebraic universes

- component: modules and universes
- introduced: a long time ago
- impacted released versions: ??-V8.8.0
- impacted coqchk versions: same
- fixed in: V8.8.1, V8.9.0 (#7798)
- found by: Gaëtan Gilbert
- GH issue number: coq/coq#7695
- exploit: see issue
- risk: needs usage of explicit algebraic universe annotations, coqchk
        may catch through defunctorialization

### Conversion machines

#### the invariant justifying some optimization was wrong for some combination of sharing side effects

- component: "lazy machine" (lazy krivine abstract machine)
- introduced: prior to V7.0
- impacted released versions: V8.0-V8.0pl4, V8.1-V8.1pl3
- impacted development branches: none
- impacted coqchk versions: ([eefe63d52](https://github.com/coq/coq/commit/eefe63d523b1b4c1b855e0f18e2574f98ff4ae64), Barras, 20 May 2008), was in beta-development for 8.2 at this time
- fixed in: master/trunk/8.2 ([f13aaec57](https://github.com/coq/coq/commit/f13aaec57df12380323edf450aec14c372422d58) / [a8b034513](https://github.com/coq/coq/commit/a8b034513e0c03ceb7e154949b15f62ac6862f3b), 15 May 2008, Barras), v8.1 ([e7611477a](https://github.com/coq/coq/commit/e7611477a0a0d1b7e8c233330def46a066985cdc), 15 May 2008, Barras), v8.0 ([6ed40a8bc](https://github.com/coq/coq/commit/6ed40a8bc000b0419f3f4731bf83d05ab5062e76), 29 Nov 2016, Herbelin, backport)
- found by: Gonthier
- exploit: by Gonthier
- GH issue number: none
- risk: unrealistic to be exploited by chance

#### collision between constructors when more than 256 constructors in a type

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- introduced: V8.1
- impacted released versions: V8.1-V8.5pl3, V8.2-V8.2pl2, V8.3-V8.3pl3, V8.4-V8.4pl5
- impacted development branches: none
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: master/trunk/v8.5 ([00894adf6](https://github.com/coq/coq/commit/00894adf6fc11f4336a3ece0c347676bbf0b4c11) / [596a4a525](https://github.com/coq/coq/commit/596a4a5251cc50f50bd6d25e36c81341bf65cfed), 26-39 Mar 2015, Grégoire), v8.4 ([cd2101a39](https://github.com/coq/coq/commit/cd2101a39b3b8d58ce569761c905a5baf1dcdc86), 1 Apr 2015, Grégoire), v8.3 ([a0c7fc05b](https://github.com/coq/coq/commit/a0c7fc05b302e38a2869c20f6db1dc376cdb59da), 1 Apr 2015, Grégoire), v8.2 ([2c6189f61](https://github.com/coq/coq/commit/2c6189f61b85bbe1a2a56754c9effc2d7a72f16d), 1 Apr 2015, Grégoire), v8.1 ([bb877e5b5](https://github.com/coq/coq/commit/bb877e5b54678bc34e4362fcf0315224e7c4f4cc), 29 Nov 2015, Herbelin, backport)
- found by: Dénès, Pédrot
- exploit: test-suite/bugs/bug_4157.v
- GH issue number: coq/coq#4157
- risk:

#### wrong universe constraints

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- introduced: possibly exploitable from V8.1; exploitable at least from V8.5
- impacted released versions: V8.1-V8.4pl5 unknown, V8.5-V8.5pl3, V8.6-V8.6.1, V8.7.0-V8.7.1
- impacted development branches: unknown for v8.1-v8.4, none from v8.5
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: master ([c9f3a6cbe](https://github.com/coq/coq/commit/c9f3a6cbe5c410256fe88580019f5c7183bab097), 12 Feb 2018, coq/coq#6713, Dénès), v8.7 ([c058a4182](https://github.com/coq/coq/commit/c058a4182b39460ba2b256c479a1389216c25ca9), 15 Feb 2018, Zimmermann, backport), v8.6 ([a2cc54c64](https://github.com/coq/coq/commit/a2cc54c649c0b13190268cc5d490342d5f0cec10), 21 Feb 2018, Herbelin, backport), v8.5 ([d4d550d0f](https://github.com/coq/coq/commit/d4d550d0f1ae5f4a8d29bbcdf991a2526ab555a6), 21 Feb 2018, Herbelin, backport)
- found by: Dénès
- exploit: test-suite/bugs/bug_6677.v
- GH issue number: coq/coq#6677
- risk:

#### missing pops in executing 31bit arithmetic

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- introduced: V8.5
- impacted released versions: V8.1-V8.4pl5
- impacted development branches: v8.1 (probably)
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: master/trunk/v8.5 ([a5e04d9dd](https://github.com/coq/coq/commit/a5e04d9dd178b2870b79776e1fbf1a858cdac49d), 6 Sep 2015, Dénès), v8.4 ([d5aa3bf6](https://github.com/coq/coq/commit/d5aa3bf6fc7382e31b6f1bac58b644843d783f13), 9 Sep 2015, Dénès), v8.3 ([5da5d751](https://github.com/coq/coq/commit/5da5d751c92df23ff3f42a04061960b287a4d3ea), 9 Sep 2015, Dénès), v8.2 ([369e82d2](https://github.com/coq/coq/commit/369e82d2cdcd0d66d0c474dc1d062a4fc62aa24a), 9 Sep 2015, Dénès),
- found by: Catalin Hritcu
- exploit: lost?
- GH issue number: ?
- risk:

#### primitive integer emulation layer on 32 bits not robust to garbage collection

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- introduced: master (before v8.10 in GH pull request coq/coq#6914)
- impacted released versions: none
- impacted development branches: v8.10
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: [5914313](https://github.com/coq/coq/commit/591431312465291e85fb352a69e947eedeb2e199) (v8.10)
- found by: Roux, Melquiond
- exploit:
- GH issue number: coq/coq#9925
- risk:

#### broken long multiplication primitive integer emulation layer on 32 bits

- component: all 3 kernel conversion machines (lazy, VM, native)
- introduced: [e43b176](https://github.com/coq/coq/commit/e43b1768d0f8399f426b92f4dfe31955daceb1a4)
- impacted released versions: 8.10.0, 8.10.1, 8.10.2
- impacted development branches: 8.11
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: [4e176a7](https://github.com/coq/coq/commit/4e176a7ee4660d505321ca55c5ce70a6c3d50d3b)
- found by: Soegtrop, Melquiond
- exploit: test-suite/bugs/bug_11321.v
- GH issue number: coq/coq#11321
- risk: critical, as any BigN computation on 32-bit architectures is wrong

#### broken addmuldiv operation for large shifts

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- impacted released versions: 8.10 to 8.19
- impacted development branches: 8.20
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: [bc0adb4](https://github.com/coq/coq/commit/bc0adb41a7c311f8d8305839c19e4812ff602720)
- found by: Martin Karup Jensen
- GH issue number: coq/coq#19402
- risk: could be exploited by chance (though not in BigNums)

#### translation of identifier from Coq to OCaml was not bijective, leading to identify True and False

For instance `α` and `__U03b1_` were the same in the native compiler.

- component: "native" conversion machine (translation to OCaml which compiles to native code)
- introduced: V8.5
- impacted released versions: V8.5-V8.5pl1
- impacted development branches: none
- impacted coqchk versions: none (no native computation in coqchk)
- fixed in: master/trunk/v8.6 ([244d7a9aa](https://github.com/coq/coq/commit/244d7a9aafe7ad613dd2095ca3126560cb3ea1d0), 19 May 2016, letouzey), v8.5 ([088b3161c](https://github.com/coq/coq/commit/088b3161c93e46ec2d865fe71a206cee15acd30c), 19 May 2016, letouzey),
- found by: Letouzey, Dénès
- exploit: see commit message for [244d7a9aa](https://github.com/coq/coq/commit/244d7a9aafe7ad613dd2095ca3126560cb3ea1d0)
- GH issue number: ?
- risk:

#### stuck primitive projections computed incorrectly by native_compute

- component: primitive projections, native_compute
- introduced: 1 Jun 2018, [e1e7888a](https://github.com/coq/coq/commit/e1e7888ac4519f4b7470cc8469f9fd924514e352), ppedrot
- impacted released versions: 8.9.0
- impacted coqchk versions: none
- fixed in: 8.9.1 coq/coq#9900
- found by: maximedenes exploiting bug coq/coq#9684
- exploit: test-suite/bugs/bug_9684.v
- GH issue number: coq/coq#9684

#### incorrect De Bruijn handling when inferring the relevance mark for a lambda

- component: lazy machine
- introduced: 2019-03-15, [23f84f37c6](https://github.com/coq/coq/commit/23f84f37c674a07e925925b7e0d50d7ee8414093) and [71b9ad8526](https://github.com/coq/coq/commit/71b9ad8526155020c8451dd326a52e391a9a8585), SkySkimmer
- impacted released versions: 8.10.0
- impacted coqchk versions: 8.10.0
- fixed in: 8.10.1 coq/coq#10904
- found by: ppedrot investigating unexpected conversion failures with SProp
- exploit: test-suite/bugs/bug_10904.v
- GH issue number: coq/coq#10904
- risk: none without using -allow-sprop (off by default in 8.10.0),
        otherwise could be exploited by mistake

#### buffer overflow on large accumulators

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- introduced: 8.1
- impacted released versions: 8.1-8.12.1
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: 8.13.0 (#13431)
- found by: Dolan, Roux, Melquiond
- GH issue number: ocaml/ocaml#6385, coq/coq#11170
- risk: medium, as it can happen for large irreducible applications

#### buffer overflow, arbitrary code execution on floating-point operations

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- introduced: 8.13
- impacted released versions: 8.13.0
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: 8.13.1
- found by: Melquiond
- GH issue number: coq/coq#13867
- risk: none, unless using floating-point operations; high otherwise;
        noticeable if activated by chance, since it usually breaks
        control-flow integrity

#### arbitrary code execution on irreducible PArray.set

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- introduced: 8.13
- impacted released versions: 8.13.0, 8.13.1
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: 8.13.2
- found by: Melquiond
- GH issue number: coq/coq#13998
- risk: none, unless using primitive array operations; systematic otherwise

#### arbitrary code execution on arrays of floating point numbers

- component: "virtual" and "native" conversion machines
- introduced: 8.13
- impacted released versions: 8.13.0, 8.13.1, 8.14.0
- impacted coqchk versions: none (no VM / native computation in coqchk)
- fixed in: 8.14.1
- found by: Melquiond
- GH issue number: coq/coq#15070
- risk: none, unless mixing open terms and primitive floats inside primitive
- arrays; critical otherwise

#### conversion of Prod / Prod values was comparing the wrong components

- component: "native" conversion machine (translation to OCaml which compiles to native code)
- introduced: V8.5
- impacted released versions: V8.5-V8.16.0 (when built with native computation enabled)
- impacted coqchk versions: none (no native computation in coqchk)
- fixed in: 8.16.1
- found by: Melquiond
- GH issue number: coq/coq#16645
- risk: systematic

#### η-expansion of cofixpoints was performed in the wrong environment

- component: "virtual" and "native" conversion machines
- introduced: V8.9
- impacted released versions: V8.9-V8.16.0
- impacted coqchk versions: none (no VM / native computation in coqchk)
- fixed in: 8.16.1
- found by: Gaëtan Gilbert and Pierre-Marie Pédrot
- GH issue number: coq/coq#16831
- risk: low, as it requires carefully crafted cofixpoints

#### conversion would compare the mutated version of primitive arrays instead of undoing mutation where needed

- component: all 3 kernel conversion machines (lazy, VM, native), primitive arrays
- introduced: V8.13
- impacted released versions: V8.13 to V8.16.0
- impacted coqchk versions: same
- fixed in: V8.16.1, V8.17
- found by: Maxime Buyse and Andres Erbsen
- exploit: Andres Erbsen
- GH issue number: coq/coq#16829
- risk: some if using primitive arrays

#### tactic code could mutate a global cache of values for section variables

- component: "virtual" reduction machine
- introduced: V8.1
- impacted released versions: V8.1-V8.16.1
- impacted coqchk versions: none (no tactics in coqchk, VM only sees checked terms)
- fixed in: V8.17.0
- found by: Gaëtan Gilbert with hint from Pierre-Marie Pédrot
- GH issue number: coq/coq#16957
- risk: the full exploitation seems to require "Definition := ltac:()"
        with change_no_check on a section variable in the ltac

#### incorrect handling of universe polymorphism

- component: VM machine
- introduced: V8.5
- impacted released versions: V8.5-V8.8.0
- impacted coqchk versions: none (no VM in coqchk)
- fixed in: V8.8.1, V8.9.0
- found by: Jason Gross
- GH issue number: coq/coq#7723
- exploit: see issue
- risk: ??

### Side-effects

#### polymorphic side-effects inside monomorphic definitions incorrectly handled as not inlined

- component: side-effects
- introduced: ?
- impacted released versions: at least from 8.6 to 8.12.0
- impacted coqchk versions: none (no side-effects in the checker)
- fixed in: V8.12.1 (coq/coq#13331)
- found by: ppedrot
- exploit: test-suite/bugs/bug_13330.v
- GH issue number: coq/coq#13330
- risk: unlikely to be exploited by mistake, requires the use of unsafe tactics

#### Section variables used in side effects not checked by Proof using

- component: side-effects / sections
- introduced: 8.11
- impacted released versions: from 8.11.0 to 8.20.0
- impacted coqchk versions: none (no side-effects in the checker)
- fixed in: V9.0 (coq/coq#19476)
- found by: Pierre Courtieu
- exploit: test-suite/output/bug_19861.v
- GH issue number: coq/coq#19861
- risk: requires incorrect `Proof using` and no use of coqchk

### Forgetting unsafe flags

#### unsafe typing flags used inside a section would not be reported by Print Assumptions after closing the section

- component: sections
- introduced: [abab878b8d](https://github.com/coq/coq/commit/abab878b8d8b5ca85a4da688abed68518f0b17bd) (#10291, 8.11), technically available earlier through plugins
- impacted coqchk versions: none (coqchk rejects affected files)
- fixed in: 8.14 coq/coq#14395
- found by: Anton Trunov
- GH issue number: coq/coq#14317
- risk: low as it needs the use of explicit unsafe flags

### Conflicts with axioms in library

#### axiom of description and decidability of equality on real numbers in library Reals was inconsistent with impredicative Set

- component: library of real numbers
- introduced: [67c75fa01](https://github.com/coq/coq/commit/67c75fa01adbbe1d4e39eff2b930ad168510072c), 20 Jun 2002
- impacted released versions: 7.3.1, 7.4
- impacted coqchk versions:
- fixed by deciding to drop impredicativity of Set: [bac707973](https://github.com/coq/coq/commit/bac707973955ef64eadae24ea01e029a5394626e), 28 Oct 2004
- found by: Herbelin & Werner
- exploit: need to find the example again
- GH issue number: no
- risk: unlikely to be exploited by chance

#### guard condition was unknown to be inconsistent with propositional extensionality in library Sets

- component: library of extensional sets, guard condition
- introduced: not a bug per se but an incompatibility discovered late
- impacted released versions: technically speaking from V6.1 with the introduction of the Sets library which was then inconsistent from the very beginning without we knew it
- impacted coqchk versions: ?
- fixed by constraining the guard condition: ([9b272a8](https://github.com/coq/coq/commit/9b272a861bc3263c69b699cd2ac40ab2606543fa), [ccd7546c](https://github.com/coq/coq/commit/ccd7546cd32c8a7901a4234f86aa23b4a7e1a043) 28 Oct 2014, Barras, Dénès)
- found by: Schepler, Dénès, Azevedo de Amorim
- exploit: ?
- GH issue number: none
- risk: unlikely to be exploited by chance (?)

#### incompatibility axiom of choice and excluded-middle with elimination of large singletons to Set

- component: library for axiom of choice and excluded-middle
- introduced: not a bug but a change of intended "model"
- impacted released versions: strictly before 8.1
- impacted coqchk versions: ?
- fixed by constraining singleton elimination: [b19397ed8](https://github.com/coq/coq/commit/b19397ed88ef8aa1ea1ca228b5d23b94e15f419f), r9633, 9 Feb 2007, Herbelin
- found by: Benjamin Werner
- exploit:
- GH issue number: none
- risk:

#### Incorrect specification of PrimFloat.leb

- component: primitive floating-points
- introduced: 8.11
- impacted released versions: 8.11.0, 8.11.1, 8.11.2
- fixed by fixing the spec: coq/coq#12484
- found by: Pierre Roux
- exploit: test-suite/bugs/bug_12483.v
- GH issue number: coq/coq#12483
- risk: proof of false when using the incorrect axiom

#### Incorrect implementation of SFclassify.

- component: floating-point library
- introduced: 8.11
- impacted released versions: 8.11.0-8.15.1
- fixed by fixing the implementation: coq/coq#16101
- found by: François Bobot
- exploit: test-suite/bugs/bug_16096.v
- GH_issue_number: coq/coq#16096
- risk: proof of false when using the axioms in Floats.Axioms.

#### nativenorm reading back closures as arbitrary floating-point values

- component: primitive floating-points + "native" conversion machine (translation to OCaml which compiles to native code)
- introduced: 8.11
- impacted released versions: 8.11.0-8.17.1
- impacted coqchk versions: none (no native computation in coqchk)
- fixed in: 8.18.0
- found by: Jason Gross
- GH issue number: coq/coq#17871
- risk: proof of false when using primitive floats and native_compute

#### guard condition issue made it inconsistent with univalence

- component: guard condition
- introduced: variant of [the issue with propositional extensionality](#guard-condition-was-unknown-to-be-inconsistent-with-propositional-extensionality-in-library-Sets) that was not fixed then
- impacted released versions: the relative inconsistency was present from the very beginning, until 9.0
- impacted coqchk versions: *-9.0
- fixed in: 9.0.1
- fixed by further constraining the guard condition: [PR 21050](https://github.com/rocq-prover/rocq/pull/21050)
- found by: Thomas Lamiaux, Yann Leray, Pierre-Marie Pédrot, Nicolas Tabareau
- GH issue number: [21053](https://github.com/rocq-prover/issues/21053)
- risk: unlikely to be exploited by chance (?)

### Deserialization

#### deserialization of .vo data not properly checked

- component: coqchk (coqc trusts that .vo files are well formed)
- introduced: 8.16 (univ levels), 8.10 (retroknowledge)
- impacted released versions: 8.10-8.18.1
- impacted coqchk versions: same
- fixed in: 8.19
- found by: Mario Carneiro
- GH issue number: N/A (fix pull requests: coq/coq#18403, coq/coq#18406)
- risk: can lead to segfaults or arbitrary code execution on crafted .vo files
    (files produced by coqc are fine)

There were otherwise several bugs in beta-releases, from memory, bugs with beta versions of primitive projections or template polymorphism or native compilation or guard (e7fc96366, 2a4d714a1).

## Probably non exploitable fixed bugs

There were otherwise maybe unexploitable kernel bugs, e.g. 2df88d83
(Require overloading), 0adf0838 ("Univs: uncovered bug in
strengthening of opaque polymorphic definitions."), 5122a398 (#3746
about functors), coq/coq#4346 (casts in VM), a14bef4 (guard condition in
8.1), 6ed40a8 ("Georges' bug" with ill-typed lazy machine), and
various other bugs in 8.0 or 8.1 without knowing if they are critical.

#### bug in 31bit arithmetic

- component: "virtual machine" (compilation to bytecode ran by a C-interpreter)
- introduced: V8.1
- impacted released versions: none
- impacted development branches:
- impacted coqchk versions: none (no virtual machine in coqchk)
- fixed in: master/trunk/v8.5 (0f8d1b92c, 6 Sep 2015, Dénès)
- found by: Dénès, from a bug report by Tahina Ramananandro
- exploit: non exploitable?
- GH issue number: ?
- risk: