| Age | Commit message (Collapse) | Author |
|---|
|
Fix bug in Shrink obligations with Program in the process.
Fix implementation of shrink for abstract proofs
- Update doc in term.mli to reflect the fact that let-in's
are part of what is returned by [decompose_lam_assum].
|
|
|
|
|
|
|
|
Cf CHANGES for details.
|
|
When typing a "with clause fails, type classes are used to possibly
help to insert coercions. If this heuristic fails, do not consider it
anymore to be the best failure since it has made type classes choices
which may be inconsistent with other constraints and on which no
backtracking is possible anymore (see new example in test suite file
4782.v).
This does not mean that using type classes at this point is good. It
may find an instance which help to find a coercion, but which might
still be a choice of instance and coercion which is incompatible with
other constraints.
I tend to think that a convenient way to go to deal with the absence
of backtracking in inserting coercions would be to have special
For the record, here is a some comments of what happens regarding
f9695eb4b and 827663982.
In the presence of an instance (x:=t) given in a "with" clause, with
t:T, and x expected of type T', the situation is the following:
Before f9695eb4b:
- If T and T' are closed and T <= T' is not satisfiable (no coercion
or not convertible), the test for possible insertion of a coercion
is postponed to w_merge, even though there is no way to get more
information since T ant T' are closed. As a result, t may be
ill-typed and the unification may try to unify ill-formed terms,
leading to #4872.
- If T and T' are not closed and contains evars of type a type class,
inference of type classes is tried. If it fails, e.g. because a
wrong type class instance is found, it was postponed to w_merge as
above, and the test for coercion is retried now interleaved with
type classes.
After f9695eb4b and 827663982e:
- If T and T' are closed and T <= T' is not satisfiable (no coercion
or not convertible), the test for possible insertion of a coercion
is an immediate failure. This fixes #4872.
- However, If T and T' are not closed and contains evars of type a
type class, inference of type classes is tried. If it gives closed
terms and fails, this is immediate failure without backtracking on
type classes, resulting in the problem added here to file 4872.v.
The current fix does not consider the result of the use of type
classes while trying to insert a coercion to be the last word on
it. So, it fails with an error which is not the error for conversion
of closed terms (ConversionFailed), therefore in a way expected by
f9695eb4b and 827663982e, and the "with" typing problem is then
postponed again.
|
|
|
|
Preserve a compatibility whether the Structural Injection flag is on
or not.
|
|
Comments
--------
- The tactic specialize conveys a somehow intuitive reasoning concept
and I would support continuing maintaining it even if the design
comes in my opinion with some oddities. (Note that the experience of
MathComp and SSReflect also suggests that specialize is an
interesting concept in itself).
There are two variants to specialize:
- specialize (H args) with H an hypothesis looks natural: we
specialize H with extra arguments and the "as pattern" clause comes
naturally as an extension of it, destructuring the result using the
pattern.
- specialize term with bindings makes the choice of fully applying the
term filling missing expressions with bindings and to then behave as
generalize. Wouldn't we like a more fine-grained approach and the
result to remain in the context?
In this second case, the "as" clause works as if the term were posed
in the context with "pose proof".
|
|
There were three versions of injection:
1. "injection term" without "as" clause:
was leaving hypotheses on the goal in reverse order
2. "injection term as ipat", first version:
was introduction hypotheses using ipat in reverse order without
checking that the number of ipat was the size of the injection
(activated with "Unset Injection L2R Pattern Order")
3. "injection term as ipat", second version:
was introduction hypotheses using ipat in left-to-right order
checking that the number of ipat was the size of the injection
and clearing the injecting term by default if an hypothesis
(activated with "Set Injection L2R Pattern Order", default one from 8.5)
There is now:
4. "injection term" without "as" clause, new version:
introducing the components of the injection in the context in
left-to-right order using default intro-patterns "?"
and clearing the injecting term by default if an hypothesis
(activated with "Set Structural Injection")
The new versions 3. and 4. are the "expected" ones in the sense that
they have the following good properties:
- introduction in the context is in the natural left-to-right order
- "injection" behaves the same with and without "as", always
introducing the hypotheses in the goal what corresponds to the
natural expectation as the changes I made in the proof scripts for
adaptation confirm
- clear the "injection" hypothesis when an hypothesis which is the
natural expectation as the changes I made in the proof scripts for
adaptation confirm
The compatibility can be preserved by "Unset Structural Injection" or
by calling "simple injection".
The flag is currently off.
|
|
simplifying and generalizing the grammar entries for injection,
discriminate and simplify_eq.
|
|
This commit documents par:, fixes its semantics so that is
behaves like all:, supports (toplevel) abstract and optimizes
toplevel solve.
`par: solve [tac]` is equivalent to `Ltac tac1 := solve[tac]...par: tac1`
but is optimized for failures: if one goal fails all are aborted
immediately.
`par: abstract tac` runs abstract on the generated proof terms. Nested
abstract calls are not supported.
|
|
along with goals, with nice formatting.
|
|
Looping on jenkins only, couldn't reproduce locally.
To be investigated further.
|
|
|
|
As noticed by C. Cohen it was confusingly different from standard
notation.
|
|
|
|
Fix test-suite files
|
|
Now that typeclasses eauto uses the new eauto.
|
|
Add an option to force backtracking at toplevel, which
is used by default when calling typeclasses eauto on a set of goals.
They might be depended on by other subgoals, so the tactic should
be backtracking by default, a once can make it not backtrack.
|
|
Report limit exceeded on _any_ branch so that we pursue search
if it was reached at least once. Add example by N. Tabareau in
test-suite.
|
|
Fix typo in proofview
|
|
with full backtracking across multiple goals.
|
|
We unify types of let-ins in FO heuristic before their bodies, using
cumulativity in either direction. This maintains the invariant that we
are comparing terms in related types throughout unification.
Also adapt test-suite file for bug #3929.
|
|
computing the arguments which allows to decide which list of implicit
arguments to consider when several such lists are available.
For instance, "eq_refl (A:=nat)" is now interpreted as "@eq_refl nat _",
the same way as if we had said:
Arguments eq_refl {A} {x}.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This is the "error resiliency" mode for STM
|
|
They can still be used at the toplevel.
|
|
This allows to write things like this:
split; 2: intro _; exact I
or like this:
eexists ?[x]; ?[x]: exact 0; trivial
This has the side-effect on making the '?' before '[x]' mandatory.
|
|
|
|
|
|
|
|
This fixes the declarations of constraints, universes
and assumptions:
- global constraints can refer to global universes only,
- polymorphic universes, constraints and assumptions can only be
declared inside sections, when all the section's
variables/universes are polymorphic as well.
- monomorphic assumptions may only be declared in section contexts
which are not parameterized by polymorphic universes/assumptions.
Add fix for part 1 of bug #4816
|
|
|
|
|
|
Check that the polymorphic status of everything that
is parameterized in nested sections is coherent.
|
|
|
|
The tentative fix in f9695eb4b (which I was afraid it might be too
strong, since it was implying failing more often) indeed broke other
things (see #4813).
|
|
Trying to now catch all unification errors, but without a clear view
at whether some errors could be tolerated at the point of checking the
type of the binding.
|
|
|
|
Typically, a problem of the form "?x args = match ?y with ... end" was
a failure even if miller-unification was applicable.
|
