diff options
author | ckeller <ckeller@users.noreply.github.com> | 2019-01-28 23:19:12 +0100 |
---|---|---|
committer | GitHub <noreply@github.com> | 2019-01-28 23:19:12 +0100 |
commit | 7021c53d4ecf97c82ccebb6bb45f5305d8b482ea (patch) | |
tree | ba7537e1e813cabf9ee0d910f845c71fa5f446e7 /README.md | |
parent | 36548d6634864a131cc83ce21491c797163de305 (diff) | |
download | smtcoq-7021c53d4ecf97c82ccebb6bb45f5305d8b482ea.tar.gz smtcoq-7021c53d4ecf97c82ccebb6bb45f5305d8b482ea.zip |
Merge from LFSC (#26)
* Showing models as coq counter examples in tactic without constructing coq terms
* also read models when calling cvc4 with a file (deactivated because cvc4 crashes)
* Show counter examples with variables in the order they are quantified in the Coq goal
* Circumvent issue with ocamldep
* fix issue with dependencies
* fix issue with dependencies
* Translation and OCaml support for extract, zero_extend, sign_extend
* Show run times of components
* print time on stdout instead
* Tests now work with new version (master) of CVC4
* fix small printing issue
* look for date on mac os x
* proof of valid_check_bbShl: some cases to prove.
* full proof of "left shift checker".
* full proof of "rigth shift checker".
* Support translation of terms bvlshr, bvshl but LFSC rules do not exists at the moment
Bug fix for bitvector extract (inverted arguments)
* Typo
* More modularity on the format of traces depending on the version of coq
* More straightforward definitions in Int63Native_standard
* Use the Int31 library with coq-8.5
* Use the most efficient operations of Int31
* Improved performance with coq-8.5
* Uniform treatment of sat and smt tactics
* Hopefully solved the problem with universes for the tactic
* Updated the installation instructions
* Holes for unsupported bit blasting rules
* Cherry-picking from smtcoq/smtcoq
* bug fix hole for bitblast
* Predefined arrays are not required anymore
* fix issue with coq bbT and bitof construction from ocaml
* bug fix in smtAtom for uninterpreted functions
fix verit test file
* fix issue with smtlib2 extract parsing
* It looks like we still need the PArray function instances for some examples (see vmcai_bytes.smt2)
* Solver specific reification:
Each solver has a list of supported theories which is passed to Atom.of_coq, this function creates uninterpreted functions / sorts for unsupported features.
* show counter-examples with const_farray instead of const for constant array definitions
* Vernacular commands to debug checkers.
Verit/Lfsc_Checker_Debug will always fail, reporting the first proof step of the certificate that failed be checked
* Update INSTALL.md
* show smtcoq proof when converting
* (Hopefully) repared the universes problems
* Corrected a bug with holes in proofs
* scripts for tests:
create a folder "work" under "lfsc/tests/", locate the benchmarks there.
create a folder "results" under "lfsc/tests/work/" in which you'll find the results of ./cvc4tocoq.
* make sure to give correct path for your benchs...
* Checker for array extensionality modulo symmetry of equality
* fix oversight with bitvectors larger than 63 bits
* some printing functions for smt2 ast
* handle smtlib2 files with more complicated equivalence with (= ... )
* revert: ./cvc4tocoq does not output lfsc proofs...
* bug fix one input was ignored
* Don't show verit translation of LFSC proof if environment variable DONTSHOWVERIT is set
(e.g. put export DONTSHOWVERIT="" in your .bashrc or .bashprofile)
* Also sort names of introduced variables when showing counter-example
* input files for which SMTCoq retuns false.
* input files for which SMTCoq retuns false.
* use debug checker for debug file
* More efficient debug checker
* better approximate number of failing step of certificate in debug checker
* fix mistake in ml4
* very first attempt to support goals in Prop
* bvs: comparison predicates in Prop and their <-> proofs with the ones in bool
farrays: equality predicate in Prop and its <-> proof with the one in bool.
* unit, Bool, Z, Pos: comparison and equality predicates in Prop.
* a typo fixed.
* an example of array equality in Prop (converted into Bool by hand)...
TODO: enhance the search space of cvc4 tactic.
* first version of cvc4' tactic: "solves" the goals in Prop.
WARNING: supports only bv and array goals and might not be complete
TODO: add support for lia goals
* cvc4' support for lia
WARNING: might not be complete!
* small fix in cvc4' and some variations of examples
* small fix + support for goals in Bool and Bool = true + use of solve tactical
WARNING: does not support UF and INT63 goals in Prop
* cvc4': better arrangement
* cvc4': Prop2Bool by context search...
* cvc4': solve tactial added -> do not modify unsolved goals.
* developer documentation for the smtcoq repo
* cvc4': rudimentary support for uninterpreted function goals in Prop.
* cvc4': support for goals with Leibniz equality...
WARNING: necessary use of "Grab Existential Variables." to instantiate variable types for farrays!
* cvc4': Z.lt adapted + better support from verit...
* cvc4': support for Z.le, Z.ge, Z.gt.
* Try arrays with default value (with a constructor for constant arrays), but extensionality is not provable
* cvc4': support for equality over uninterpreted types
* lfsc demo: goals in Coq's Prop.
* lfsc demo: goals in Bool.
* Fix issue with existential variables generated by prop2bool.
- prop2bool tactic exported by SMTCoq
- remove useless stuff
* update usage and installation instructions
* Update INSTALL.md
* highlighting
* the tactic: bool2prop.
* clean up
* the tactic smt: very first version.
* smt: return unsolved goals in Prop.
* Show when a certificate cannot be checked when running the tactic instead of at Qed
* Tactic improvements
- Handle negation/True/False in prop/bool conversions tactic.
- Remove alias for farray (this caused problem for matching on this type in tactics).
- Tactic `smt` that combines cvc4 and veriT.
- return subgoals in prop
* test change header
* smt: support for negated goals + some reorganization.
* conflicts resolved + some reorganization.
* a way to solve the issue with ambiguous coercions.
* reorganization.
* small change.
* another small change.
* developer documentation of the tactics.
* developer guide: some improvements.
* developer guide: some more improvements.
* developer guide: some more improvements.
* developer guide: some more improvements.
* pass correct environment for conversion + better error messages
* cleaning
* ReflectFacts added.
* re-organizing developers' guide.
* re-organizing developers' guide.
* re-organizing developers' guide.
* removing unused maps.
* headers.
* artifact readme getting started...
* first attempt
* second...
* third...
* 4th...
* 5th...
* 6th...
* 7th...
* 8th...
* 9th...
* 10th...
* 11th...
* 12th...
* 13th...
* 14th...
* 15th...
* 16th...
* 17th...
* Update artifact.md
Use links to lfsc repository like in the paper
* 18th...
* 19th...
* 20th...
* 21st...
* 22nd...
* 23rd...
* 24th...
* 25th...
* 26th...
* 27th...
* 28th...
* Update artifact.md
Small reorganization
* minor edits
* More minor edits
* revised description of tactics
* Final pass
* typo
* name changed: artifact-readme.md
* file added...
* passwd chaged...
* links...
* removal
* performance statement...
* typos...
* the link to the artifact image updated...
* suggestions by Guy...
* aux files removed...
* clean-up...
* clean-up...
* some small changes...
* small fix...
* additional information on newly created files after running cvc4tocoq script...
* some small fix...
* another small fix...
* typo...
* small fix...
* another small fix...
* fix...
* link to the artifact image...
* We do not want to force vm_cast for the Theorem commands
* no_check variants of the tactics
* TODO: a veriT test does not work anymore
* Compiles with both versions of Coq
* Test of the tactics in real conditions
* Comment on this case study
* an example for the FroCoS paper.
* Fix smt tactic that doesn't return cvc4's subgoals
* readme modifications
* readme modifications 2
* small typo in readme.
* small changes in readme.
* small changes in readme.
* typo in readme.
* Sync with https://github.com/LFSC/smtcoq
* Port to Coq 8.6
* README
* README
* INSTALL
* Missing file
* Yves' proposition for installation instructions
* Updated link to CVC4
* Compiles again with native-coq
* Compiles with both versions of Coq
* Command to bypass typechecking when generating a zchaff theorem
* Solved bug on cuts from Hole
* Counter-models for uninterpreted sorts (improves issue #13)
* OCaml version note (#15)
* update .gitignore
* needs OCaml 4.04.0
* Solving merge issues (under progress)
* Make SmtBtype compile
* Compilation of SmtForm under progress
* Make SmtForm compile
* Make SmtCertif compile
* Make SmtTrace compile
* Make SatAtom compile
* Make smtAtom compile
* Make CnfParser compile
* Make Zchaff compile
* Make VeritSyntax compile
* Make VeritParser compile
* Make lfsc/tosmtcoq compile
* Make smtlib2_genconstr compile
* smtCommand under progress
* smtCommands and verit compile again
* lfsc compiles
* ml4 compiles
* Everything compiles
* All ZChaff unit tests and most verit unit tests (but taut5 and un_menteur) go through
* Most LFSC tests ok; some fail due to the problem of verit; a few fail due to an error "Not_found" to investigate
* Authors and headings
* Compiles with native-coq
* Typo
Diffstat (limited to 'README.md')
-rw-r--r-- | README.md | 149 |
1 files changed, 124 insertions, 25 deletions
@@ -5,11 +5,13 @@ SAT and SMT solvers. It relies on a certified checker for such witnesses. On top of it, vernacular commands and tactics to interface with the SAT solver zChaff -and the SMT solver veriT are provided. It is designed in a modular way +and the SMT solvers veriT and CVC4 are provided. It is designed in a modular way allowing to extend it easily to other solvers. +<!--- Extraction is probably broken SMTCoq also provides an extracted version of the checker, that can be run outside Coq. +---> The current stable version is version 1.3. @@ -33,34 +35,41 @@ easily re-usable for your own usage. #### Overview -The SMTCoq module can be used in Coq files via the `Require Import -SMTCoq.` command. For each supported solver, it provides: +After installation, the SMTCoq module can be used in Coq files via the +`Require Import SMTCoq.` command. For each supported solver, it +provides: -- a vernacular command to check answers: - `XXX_Checker "problem_file" "witness_file"` returns `true` only if - `witness_file` contains a zChaff proof of the unsatisfiability of the - problem stated in `problem_file`; +- a vernacular command to check answers: `XXX_Checker "problem_file" + "witness_file"` returns `true` only if `witness_file` contains a proof + of the unsatisfiability of the problem stated in `problem_file`; - a vernacular command to safely import theorems: `XXX_Theorem theo "problem_file" "witness_file"` produces a Coq term - `teo` whose type is the theorem stated in `problem_file` if + `theo` whose type is the theorem stated in `problem_file` if `witness_file` is a proof of the unsatisfiability of it, and fails otherwise. -- a safe tactic to try to solve a Coq goal using the chosen solver. +- safe tactics to try to solve a Coq goal using the chosen solver (or a + combination of solvers). +<!--- Extraction is probably broken The SMTCoq checker can also be extracted to OCaml and then used independently from Coq. +---> +We now give more details for each solver. + +<!--- Extraction is probably broken We now give more details for each solver, and explanations on extraction. +---> #### zChaff Compile and install zChaff as explained in the installation instructions. In the following, we consider that the command `zchaff` is -in your `PATH` variable environment. +in your `PATH` environment variable. ##### Checking zChaff answers of unsatisfiability and importing theorems @@ -72,7 +81,7 @@ To check the result given by zChaff on an unsatisfiable dimacs file produces a proof witness file named `resolve_trace`. - In a Coq file `file.v`, put: -``` +```coq Require Import SMTCoq. Zchaff_Checker "file.cnf" "resolve_trace". ``` @@ -82,7 +91,7 @@ Zchaff_Checker "file.cnf" "resolve_trace". - You can also produce Coq theorems from zChaff proof witnesses: the commands -``` +```coq Require Import SMTCoq. Zchaff_Theorem theo "file.cnf" "resolve_trace". ``` @@ -93,17 +102,18 @@ will produce a Coq term `theo` whose type is the theorem stated in ##### zChaff as a Coq decision procedure The `zchaff` tactic can be used to solve any goal of the form: -``` +```coq forall l, b1 = b2 ``` -where `l` is a list of Booleans (that can be concrete terms). +where `l` is a quantifier-free list of variables and `b1` and `b2` are +expressions of type `bool`. #### veriT Compile and install veriT as explained in the installation instructions. In the following, we consider that the command `veriT` is in your `PATH` -variable environment. +environment variable. ##### Checking veriT answers of unsatisfiability and importing theorems @@ -112,13 +122,13 @@ To check the result given by veriT on an unsatisfiable SMT-LIB2 file `file.smt2`: - Produce a veriT proof witness: -``` +```coq veriT --proof-prune --proof-merge --proof-with-sharing --cnf-definitional --disable-e --disable-ackermann --input=smtlib2 --proof=file.log file.smt2 ``` This command produces a proof witness file named `file.log`. - In a Coq file `file.v`, put: -``` +```coq Require Import SMTCoq. Section File. Verit_Checker "file.smt2" "file.log". @@ -128,9 +138,9 @@ End File. - Compile `file.v`: `coqc file.v`. If it returns `true` then veriT indeed proved that the problem was unsatisfiable. -- You can also produce Coq theorems from zChaff proof witnesses: the +- You can also produce Coq theorems from veriT proof witnesses: the commands -``` +```coq Require Import SMTCoq. Section File. Verit_Theorem theo "file.smt2" "file.log". @@ -139,16 +149,105 @@ End File. will produce a Coq term `theo` whose type is the theorem stated in `file.smt2`. -The theories that are currently supported are `QF_UF`, `QF_LIA`, -`QF_IDL` and their combinations. +The theories that are currently supported by these commands are `QF_UF` +(theory of equality), `QF_LIA` (linear integer arithmetic), `QF_IDL` +(integer difference logic), and their combinations. ##### veriT as a Coq decision procedure -The `verit` tactic can be used to solve any goal of the form: +The `verit_bool` tactic can be used to solve any goal of the form: +```coq +forall l, b1 = b2 +``` +where `l` is a quantifier-free list of variables and `b1` and `b2` are +expressions of type `bool`. + +In addition, the `verit` tactic applies to Coq goals of sort `Prop`: it +first converts the goal into a term of type `bool` (thanks to the +`reflect` predicate of `SSReflect`), and then calls the previous tactic +`verit_bool`. + +The theories that are currently supported by these tactics are `QF_UF` +(theory of equality), `QF_LIA` (linear integer arithmetic), `QF_IDL` +(integer difference logic), and their combinations. + + +#### CVC4 + +Compile and install `CVC4` as explained in the installation +instructions. In the following, we consider that the command `cvc4` is +in your `PATH` environment variable. + + +##### Checking CVC4 answers of unsatisfiability and importing theorems + +To check the result given by CVC4 on an unsatisfiable SMT-LIB2 file +`name.smt2`: + +- Produce a CVC4 proof witness: + +```bash +cvc4 --dump-proof --no-simplification --fewer-preprocessing-holes --no-bv-eq --no-bv-ineq --no-bv-algebraic name.smt2 > name.lfsc ``` + +This set of commands produces a proof witness file named `name.lfsc`. + +- In a Coq file `name.v`, put: +```coq +Require Import SMTCoq Bool List. +Import ListNotations BVList.BITVECTOR_LIST FArray. +Local Open Scope list_scope. +Local Open Scope farray_scope. +Local Open Scope bv_scope. + +Section File. + Lfsc_Checker "name.smt2" "name.lfsc". +End File. +``` + +- Compile `name.v`: `coqc name.v`. If it returns `true` then the problem + is indeed unsatisfiable. + +NB: Use `cvc4tocoq` script in `src/lfsc/tests` to automatize the above steps. + +- Ex: `./cvc4tocoq name.smt2` returns `true` only if the problem + `name.smt2` has been proved unsatisfiable by CVC4. + +The theories that are currently supported by these commands are `QF_UF` +(theory of equality), `QF_LIA` (linear integer arithmetic), `QF_IDL` +(integer difference logic), `QF_BV` (theory of fixed-size bit vectors), +`QF_A` (theory of arrays), and their combinations. + + +##### CVC4 as a Coq decision procedure + +The `cvc4_bool` tactic can be used to solve any goal of the form: +```coq forall l, b1 = b2 ``` -where `l` is a list of Booleans. Those Booleans can be any concrete -terms. The theories that are currently supported are `QF_UF`, `QF_LIA`, -`QF_IDL` and their combinations. + +where `l` is a quantifier-free list of variables and `b1` and `b2` are +expressions of type `bool`. + +In addition, the `cvc4` tactic applies to Coq goals of sort `Prop`: it + first converts the goal into a term of type `bool` (thanks to the + `reflect` predicate of `SSReflect`), it then calls the previous tactic + `cvc4_bool`, and it finally converts any unsolved subgoals returned by + CVC4 back to `Prop`, thus offering to the user the possibility to solve + these (usually simpler) subgoals. + +The theories that are currently supported by these tactics are `QF_UF` +(theory of equality), `QF_LIA` (linear integer arithmetic), `QF_IDL` +(integer difference logic), `QF_BV` (theory of fixed-size bit vectors), +`QF_A` (theory of arrays), and their combinations. + + +### The smt tactic + +The more powerful tactic `smt` combines all the previous tactics: it +first converts the goal to a term of type `bool` (thanks to the +`reflect` predicate of `SSReflect`), it then calls a combination of the +`cvc4_bool` and `verit_bool` tactics, and it finally converts any +unsolved subgoals back to `Prop`, thus offering to the user the +possibility to solve these (usually simpler) subgoals. |