| Age | Commit message (Collapse) | Author |
|---|
|
|
|
The test is refined to handle aliases: i.e. undefined evars coming from
restrictions and evar-evar unifications with an initial evar are not
considered fresh unresolved evars. To check this, we generalize the
restricted_evars set to an aliased_evars set in the evar map,
registering evars being solved by another evar due to restriction
or evar-evar unifications. This implements the proposal of PR #370
for testing the resolution status of evars independently of the evar-evar
orientation order.
This allows [apply] to refine an evar with a new one if it results from a
[clear] request or an evar-evar solution only, otherwise the new evar is
considered fresh and an error is raised.
Also fixes bugs #4095 and #4413.
Co-authored-by: Maxime Dénès <maxime.denes@inria.fr>
|
|
Before this patch, the proof engine had three notions of shelves:
- A local shelf in `proofview`
- A global shelf in `Proof.t`
- A future shelf in `evar_map`
This has lead to a lot of confusion and limitations or bugs, because
some components have only a partial view of the shelf: the pretyper can
see only the future shelf, tactics can see only the local and future
shelves. In particular, this refactoring is needed for #7825.
The solution we choose is to move shelf information to the evar map, as
a shelf stack (for nested `unshelve` tacticals).
Closes #8770.
Closes #6292.
Co-authored-by: Gaëtan Gilbert <gaetan.gilbert@skyskimmer.net>
|
|
|
|
|
|
|
|
We try to encapsulate the future goals abstraction in the evar map.
A few calls to `save_future_goals` and `restore_future_goals` are still
there, but we try to minimize them.
This is a preliminary refactoring to make the invariants between the
shelf and future goals more explicit, before giving unification access
to the shelf, which is needed for #7825.
|
|
Current backtraces for tactics leave a bit to desire, for example
given the program:
```coq
Lemma u n : n + 0 = n.
rewrite plus_O_n.
```
the backtrace stops at:
```
Found no subterm matching "0 + ?M160" in the current goal.
Called from file "proofs/proof.ml", line 381, characters 4-42
Called from file "tactics/pfedit.ml", line 102, characters 31-58
Called from file "plugins/ltac/g_ltac.mlg", line 378, characters 8-84
```
Backtrace information `?info` is as of today optional in some tactics,
such as `tclZERO`, it doesn't cost a lot however to reify backtrace
information indeed in `tclZERO` and provide backtraces for all tactic
errors. The cost should be small if we are not in debug mode.
The backtrace for the failed rewrite is now:
```
Found no subterm matching "0 + ?M160" in the current goal.
Raised at file "pretyping/unification.ml", line 1827, characters 14-73
Called from file "pretyping/unification.ml", line 1929, characters 17-53
Called from file "pretyping/unification.ml", line 1948, characters 22-72
Called from file "pretyping/unification.ml", line 2020, characters 14-56
Re-raised at file "pretyping/unification.ml", line 2021, characters 66-73
Called from file "proofs/clenv.ml", line 254, characters 12-58
Called from file "proofs/clenvtac.ml", line 95, characters 16-53
Called from file "engine/proofview.ml", line 1110, characters 40-46
Called from file "engine/proofview.ml", line 1115, characters 10-34
Re-raised at file "clib/exninfo.ml", line 82, characters 4-38
Called from file "proofs/proof.ml", line 381, characters 4-42
Called from file "tactics/pfedit.ml", line 102, characters 31-58
Called from file "plugins/ltac/g_ltac.mlg", line 378, characters 8-84
```
which IMO is much better.
|
|
Add headers to a few files which were missing them.
|
|
Check that we don't regress on PR #10762 example
Fix regression discovered by Arthur in PR #10762
Fix script of #10298 which was relying on breaking semantics for `eapply`
Add doc
Add comment in clenvtac
Actually, always mark shelved goals as unresolvable
Update doc to reflect semantics w.r.t. shelved subgoals
|
|
|
|
We move the role data into the evarmap instead.
|
|
We remove unused parts of the API, almost all of them belonging to the
legacy engine. This was detected using coverage testing.
The list is provisional and should be subject to change, let's see
what CI says.
|
|
|
|
We remove all calls to `Flags.is_program_mode` except one (to compute
the default value of the attribute). Everything else is passed
explicitely, and we remove the special logic in the interpretation loop
to set/unset the flag.
This is especially important since the value of the flag has an impact on
proof modes, so on the separation of parsing and execution phases.
|
|
This should improve correctness and will be needed for the PRs that
remove global access to the proof state.
|
|
This is a pre-requisite to use automated formatting tools such as
`ocamlformat`, also, there were quite a few places where the comments
had basically no effect, thus it was confusing for the developer.
p.s: Reading some comments was a lot of fun :)
|
|
This avoids all the side effects associated with the manipulation of an
unresolvable flag. In the new design:
- The evar_map stores a set of evars that are candidates for typeclass
resolution, which can be retrieved and set.
We maintain the invariant that it always contains only undefined
evars.
- At the creation time of an evar (new_evar), we classify it as a
potential candidate of resolution.
- This uses a hook to test if the conclusion ends in a typeclass
application. (hook set in typeclasses.ml)
- This is an approximation if the conclusion is an existential (i.e.
not yet determined). In that case we register the evar as
potentially a typeclass instance, and later phases must consider
that case, dropping the evar if it is not a typeclass.
- One can pass the ~typeclass_candidate:false flag to new_evar to
prevent classification entirely. Typically this is for new goals
which should not ever be considered to be typeclass resolution
candidates.
- One can mark a subset of evars unresolvable later if
needed. Typically for clausenv, and marking future goals as
unresolvable even if they are typeclass goals. For clausenv for
example, after turing metas into evars we first (optionally) try a
typeclass resolution on the newly created evars and only then mark
the remaining newly created evars as subgoals. The intent of the
code looks clearer now.
This should prevent keeping testing if undefined evars are classes
all the time and crawling large sets when no typeclasses are present.
- Typeclass candidate evars stay candidates through
restriction/evar-evar solutions.
- Evd.add uses ~typeclass_candidate:false to avoid recomputing if the new
evar is a candidate. There's a deficiency in the API, in most use
cases of Evd.add we should rather use a:
`Evd.update_evar_info : evar_map -> Evar.t -> (evar_info -> evar_info)
-> evar_map`
Usually it is only about nf_evar'ing the evar_info's contents, which
doesn't change the evar candidate status.
- Typeclass resolution can now handle the set of candidates
functionally: it always starts from the set of candidates (and not the
whole undefined_map) and a filter on it, potentially splitting it in
connected components, does proof search for each component in an
evar_map with an empty set of typeclass evars (allowing clean
reentrancy), then reinstates the potential remaining unsolved
components and filtered out typeclass evars at the end of
resolution.
This means no more marking of resolvability/unresolvability
everywhere, and hopefully a more efficient implementation in general.
- This is on top of the cleanup of evar_info's currently but can
be made independent.
[typeclasses] Fix cases.ml: none of the new_evars should be typeclass candidates
Solve bug in inheritance of flags in evar-evar solutions.
Renaming unresolvable to typeclass_candidate (positive) and fix maybe_typeclass_hook
|
|
|
|
We remove internal functions and types from the API.
|
|
We simply exploit a type isomorphism to remove the use of dedicated algebraic
types in the kernel which are actually not necessary.
|
|
|
|
We bootstrap the circular evar_map <-> econstr dependency by moving
the internal EConstr.API module to Evd.MiniEConstr. Then we make the
Evd functions use econstr.
|
|
Tactic-in-term can be called from within a tactic itself. We have to
preserve the preexisting future_goals (if called from pretyping) and
we have to inform of the existence of pending goals, using
future_goals which is the only way to tell it in the absence of being
part of an encapsulating proofview.
This fixes #6313.
Conversely, future goals, created by pretyping, can call ltac:(giveup) or
ltac:(shelve), and this has to be remembered. So, we do it.
|
|
|
|
|
|
|
|
|
|
|
|
In current code, `Proofview.Goal.t` uses a phantom type to indicate
whether the goal was properly substituted wrt current `evar_map` or
not.
After the introduction of `EConstr`, this distinction should have
become unnecessary, thus we remove the phantom parameter from
`'a Proofview.Goal.t`. This may introduce some minor incompatibilities
at the typing level. Code-wise, things should remain the same.
We thus deprecate `assume`. In a next commit, we will remove
normalization as much as possible from the code.
|
|
Because the argument given to refine may mess with the evarmap, the goal being
refined can be solved by side-effect after the term filler is computed. If this
happens, we simply don't perform the refining operation.
|
|
|
|
|
|
|
|
Reminder of (some of) the reasons for removal:
- Despite the claim in sigma.mli, it does *not* prevent evar
leaks, something like:
fun env evd ->
let (evd',ev) = new_evar env evd in
(evd,ev)
will typecheck even with Sigma-like type annotations (with a proof of
reflexivity)
- The API stayed embryonic. Even typing functions were not ported to
Sigma.
- Some unsafe combinators (Unsafe.tclEVARS) were replaced with slightly
less unsafe ones (e.g. s_enter), but those ones were not marked unsafe
at all (despite still being so).
- There was no good story for higher order functions manipulating evar
maps. Without higher order, one can most of the time get away with
reusing the same name for the updated evar map.
- Most of the code doing complex things with evar maps was using unsafe
casts to sigma. This code should be fixed, but this is an orthogonal
issue.
Of course, this was showing a nice and elegant use of GADTs, but the
cost/benefit ratio in practice did not seem good.
|
|
|
|
|
|
Now it is a private field, locations are optional.
|
|
|
|
|
|
|
|
|
|
This allows to factorize code and prevents the unnecessary use of back and
forth conversions between the various types of terms.
Note that functions from typing may now raise errors as PretypeError rather
than TypeError, because they call the proper wrapper. I think that they were
wrongly calling the kernel because of an overlook of open modules.
|
|
|
|
|
|
|
|
|
|
|
|
reconsider_conv_pbs -> reconsider_unif_constraints
consider_remaining_unif_problems -> solve_unif_constraints_with_heuristics
|
|
Refine fix for bug #4763, fixing #5149
Tactic [Refine.solve_constraints] and global option
Adds a new multi-goal tactic [Refine.solve_constraints] that forces solving of
unification constraints and evar candidates to be solved. run_tactic now calls
[solve_constraints] at every [.], preserving (mostly) the 8.4/8.5 behavior of tactics.
The option allows to unset the forced solving unification constraints at
each ".", letting the user control the places where the use of
heuristics is done.
Fix test-suite files too.
|
