| Age | Commit message (Collapse) | Author |
|---|
|
vernac/
Ack-by: Janno
Reviewed-by: SkySkimmer
|
|
Reviewed-by: SkySkimmer
Reviewed-by: ppedrot
|
|
This is useful to remove some duplicate bits in other declare files.
|
|
|
|
Patch suggested by Gaëtan Gilbert
|
|
The code is self-contained and only used by commands; this also
highlights the several `Libobject.obj` registered for each
declaration.
|
|
The code is self-contained and only used by commands; this also
highlights the several `Libobject.obj` registered for each
declaration.
|
|
|
|
There are no more users.
|
|
|
|
|
|
Ack-by: ejgallego
Ack-by: gares
Reviewed-by: ppedrot
|
|
Reviewed-by: SkySkimmer
Reviewed-by: gares
|
|
Reviewed-by: JasonGross
Reviewed-by: Zimmi48
Reviewed-by: ejgallego
|
|
|
|
This allows UI clients to implement a different state management
strategy with regards to proofs, and in particular to override
`Vernacinterp.interp`.
This is work in progress towards having a true `VtTactic` which shall
not perform any state changes non-functionally, and actually removing
the series of `assert false` due to meta-vernacs.
|
|
|
|
The general idea is to move tests on scope=Discharge and on
Lib.sections_are_opened up in the call stack. This allows better
control over the universe manipulation.
There are some corner case behaviour change, eg:
- [Context (foo:=bla)] outside a section correctly declares an
axiom (fix #10668)
- (not observable) universes for [Variable A B : Type] in section are declared only once
- universes and universe names for [Axiom A B : Type] are declared
only once. This changes the qualification of the universe name:
before it was the last axiom (so [B.u0]), now it's the first
one ([A.u0]).
Probably nobody cares about this.
- context outside section uses different [kind]
I'm not sure why context outside a section behaves differently based
on whether we're in a module type, I tried to preserve it but maybe we
should uniformize.
The universe manipulation for Axiom (in the declare_assumptions
function) is a bit awkward, maybe when there are multiple monomorphic
axioms instead of trying to attach the universes to the first one we
should just declare them separately like with Context. OTOH unlike
with context dropping the universe names seems incorrect.
|
|
Primitives don't have anything to do with assumptions.
|
|
(letins still declare universes in declare_variable as they use
entries)
The section check_same_poly is moved to declare_universe_context (it
makes more sense there, universe polymorphism doesn't apply to the
variables/letins themselves)
|
|
We only do it for entries and not declarations because the upper layers
rely on the kernel being able to quickly tell that a definition is improperly
used inside a section. Typically, tactics can mess with the named context
and thus make the use of section definitions illegal. This cannot happen in
the kernel but we cannot remove it due to the code dependency.
Probably fixing a soundness bug reachable via ML code only. We were doing
fancy things w.r.t. computation of the transitive closure of the the variables,
in particular lack of proper sanitization of the kernel input.
|
|
sections
Reviewed-by: ppedrot
|
|
Ack-by: SkySkimmer
Reviewed-by: herbelin
Reviewed-by: ppedrot
|
|
We disallow adding univ constraints wich refer to polymorphic
universes, and monomorphic constants and inductives when polymorphic
universes or constraints are present.
Every other combination is already correctly discharged by the kernel.
|
|
Ack-by: SkySkimmer
Reviewed-by: ppedrot
|
|
Reviewed-by: mattam82
|
|
Now that we place imperative module declaration on top of module
interpretation we can remove the abstraction layer used in
`Declaremods`, so the `interp_modast` parameter goes away.
Improvement suggested by Gaëtan Gilbert.
|
|
We move `Declaremods` to the vernac layer as it implement
vernac-specific logic to manipulate modules which moreover is highly
imperative.
This forces code [such as printing] to manipulate the _global
imperative_ state which is a bit fishy.
The key improvement in this PR is that now `Global` is not used
anymore in `library`, so we can proceed to move it upwards.
This move is a follow-up of #10562 and a step towards moving `Global`
upper, likely to `interp` in the short term.
|
|
This helps extraction by not building sigT which can lower to Prop by
template polymorphism.
Bug #10757 can probably still be triggered by using module functors to
hide that we're using Prop from Program Fixpoint but that's probably
unfixable without fixing extraction vs template polymorphism in
general.
In passing we notice that Program doesn't know how to telescope SProp
arguments, we would need a bunch of variants of sigma types to deal
with it (or use Box?) so let's figure it out some other time.
We also reuse the universe instance to avoid generating a bunch of
short-lived universes in the universe polymorphic case.
|
|
|
|
`Import` does not actually need to register an object, only `Export`
does. So we specialize and rename the object into `ExportObject`.
|
|
Libraries are now handled like other modules.
|
|
Ack-by: ejgallego
Reviewed-by: gares
|
|
Reviewed-by: maximedenes
|
|
Ack-by: JasonGross
Ack-by: SkySkimmer
Ack-by: Zimmi48
Ack-by: herbelin
Ack-by: mattam82
Reviewed-by: ppedrot
|
|
This is step 1 on removing library state from the lower layers.
Here we move library loading to the vernacular layer; few things
depend on it:
- printers: we add a parameter for those needing to access on-disk data,
- coqlib: indeed a few tactics do try to check that a particular
library is loaded; this is a tricky part. I've replaced that for a
module name check, but indeed this is fully equivalent due to
side-effects of `Require`. We may want to think what to do here.
A few other minor code movements were needed, but there are
self-explanatory.
|
|
proof data on top of declare.
Reviewed-by: ppedrot
|
|
As documented in the feedback API.
|
|
Non-delayed entries can be done with the current constructor, delayed
ones will require more work.
|
|
top of declare.
This PR is a follow up to #10406 , moving the then introduced
`proof_entry` type to `Declare`.
This makes sense as `Declare` is the main consumer of the entry type,
and already provides the constructors for it.
This is a step towards making the entry type private, which will allow
us to enforce / handle invariants on entry data better.
A side-effect of this PR is that now `Proof_global` does depend on
`Declare`, not the other way around, but that makes sense given that
closing an interactive proof will be a client of declare.
Indeed, all `Declare` / `Pfedit` / and `Proof_global` are tied into
tactics due to `abstract`, at some point we may be able to unify all
them into a single file in `vernac`.
|
|
Fix changelog entry
Fix build of the user manual
Markup fixes from Théo Zimmermann
Update doc and changelog and improve error messages.
|
|
This could be Prop (for compat with usual Coq), Set (for HoTT),
or actually an arbitrary "i".
Take lower bound of universes into account in pretyping/engine
Reinstate proper elaboration of SProp <= l constraints:
replacing is_small with equality with lbound is _not_ semantics preserving!
lbound = Set
Elaborate template polymorphic inductives with lower bound Prop
This will make more constraints explicit
Check univ constraints with Prop as lower bound for template inductives
Restrict template polymorphic universes to those not bounded from below
Fixes #9294
fix suggested by Matthieu
Try second fix suggested by Matthieu
Take care of modifying elaboration for record declarations as well.
Rebase and export functions for debug
Remove exported functions used while debugging
Add a new typing flag "check_template" and option "-no-template-checl"
This parameterizes the new criterion on template polymorphic inductives
to allow bypassing it (necessary for backward compatibility).
Update checker to the new typing flags structure
Switch on the new template_check flag to allow old unsafe behavior in
indTyping.
This is the only change of code really impacting the kernel, together
with the commit implementing unbounded from below and parameterization
by the lower bound on universes.
Add deprecated option `Unset Template Check` allowing to make proof
scripts work with both 8.9 and 8.10 for a while
Fix `Template Check` option name and test it
Add `Unset Template Check` to Coq89.v
Cooking of inductives and template-check tests
Cleanup test-suite file for template check / universes(template) flags
cookind tests
Move test of `Unset Template Check` to the failure/ dir, but comment it
for now
Template test-suite test explanation
Overlays for PR 9918
Overlay for paramcoq
Add overlay for fiat_parsers (-no-template-check)
Add overlay for fiat_crypto_legacy
Update fiat-crypto legacy overlay
Now it points at the version that I plan on merging; I am hoping that doing this will guard against mistakes by adding an extra check that the target tested by Coq's CI on this branch works with the change I made.
Remove overlay that should no longer be necessary
The setting in the compat file should handle it
Remove now-merged fiat-crypto-legacy overlay
Update `Print Assumptions` to reflect the typing flag for template checking
Fix About and Print Assumptions for template poly, giving info on which
variables are actually polymorphic
Fix pretty printing to print global universe levels properly
Fix printing of template polymorphic universes
Fix pretty printing for template polymorphism on no universe
Fix interaction of template check and universes(template) flag
Fix indTyping to really check if there is any point in polymorphism: the
conclusion sort should be parameterized over at least one local universe
Indtyping fixes for template polymorphic Props
Allow explicit template polymorphism again
Adapt to new indTyping interface
Handle the case of template-polymorphic on no universes
correctly (morally Type0m univ represented as Prop).
Fix check of meaningfullness of template polymorphism in the kernel.
It is now done w.r.t the min_univ, the minimal universe inferred for the
inductive/record type, independently of the user-written annotation
which must only be larger than min_univ. This preserves compatibility
with UniMath and template-polymorphism as it has been implemented up-to
now.
Comment on identity non-template-polymorphism
Remove incorrect universes(template) attributes from ssr
simpl_fun can be meaningfully template-poly, as well as
pred_key (although the use is debatable: it could just
as well be in Prop).
Move `fun_of_simpl` coercion declaration out of section to respect
uniform inheritance
Remove incorrect uses of #[universes(template)] from the stdlib
Extraction of micromega changes due to moving an ind decl out of a section
Remove incorrect uses of #[universes(template)] from plugins
Fix test-suite files, removing incorrect #[universes(template)] attributes
Remove incorrect #[universes(template)] attributes in test-suite
Fix test-suite
Remove overlays as they have been merged upstream.
|
|
Fixes #10640
We remove the `StdOut` dump target, so now dump will only happen if a
file is specified. Indeed, we make the default no to dump, and enable
dump only in coqc, moving the option to the `Coqcargs` module.
No need for a changes entry as this feature was undocumented, and no
use case was given when introduced.
Output to feedback must be explicitly enabled by clients / coqidetop,
and we have thus also removed the undocumented option `-feedback-glob`.
|
|
This improves error reporting. Addendum to #10515
|
|
We remove calls of `Lemmas.Info.make` that where using the default
parameters, as this is mostly dead code now.
This brings into question quite a few things, in particular, the
uneven support of `scope` attributes by different commands / plugins.
We don't attempt to solve that yet, hopefully the ongoing constant
saving path refactoring will be able to take care of these
inconsistencies.
|
|
Reviewed-by: SkySkimmer
|
|
Reviewed-by: SkySkimmer
Reviewed-by: Zimmi48
|
|
|
|
We move `binder_kind` to the pretyping AST, removing the last data
type in the now orphaned file `Decl_kinds`.
This seems a better fit, as this data is not relevant to the lower
layers but only used in `Impargs`.
We also move state keeping to `Impargs`, so now implicit declaration
must include the type. We also remove a duplicated function.
|
|
type-in-type universes
|
