| Age | Commit message (Collapse) | Author |
|---|
|
|
|
We introduce the `with_lock` combinator that locks a mutex in an atomic fashion.
This ensures that exceptions thrown by signals will not leave the system in a
deadlocked state.
|
|
This tends to confuse the OCaml compiler, for good reasons. Indeed, if
there are mutable fields, the generated code cannot wait for the function
to be fully applied. It needs to recover the value of the mutable fields
as early as possible, and thus to create a closure.
Example:
let foo {bar} x = ...
is compiled as
let foo y = match y with {bar} -> fun x -> ...
|
|
|
|
We also spill (some) non-generic arguments and initialization code
out of coqargs and to coqtop, namely colors for the terminal. There are
more of these, left to later commits.
|
|
This component holds the code for initializing Coq:
- parsing arguments not specific to the toplevel
- initializing all components from vernac downwards (no stm)
This commit moves stm specific arguments parsing to stm/stmargs.ml
|
|
|
|
|
|
to include most common reason when this happens on accident
|
|
After the previous commit, the stm should correctly pass proof
information, thus we make `proof_object` carry it removing a bunch of
internal code.
|
|
Since 8.5, the STM could not delegate proof information it was
contained inside a closure. This could potentially create some
problems, as witnessed in #12586.
Recent refactoring have reified however much of this state, so it
seems a good idea to use bits of the state now, at the cost of
introducing some safeguards in `declare.ml` w.r.t. `Ephemerons`.
|
|
|
|
The current implementation of par: is still in the STM, but is optional.
If the STM does not take over it, it defaults to the implementation of
in comTactic which is based on all: (i.e. sequential).
This commit also moved the interpretation of a tactic from g_ltac to
vernac/comTactic which is more appropriate.
Co-authored-by: Gaëtan Gilbert <gaetan.gilbert@skyskimmer.net>
|
|
- take just a ugraph instead of the whole env
- rename to update_sigma_univs
- push global env lookup a bit further up
- fix vernacinterp call to update all surrounding proofs, not just the
top one
- flip argument order for nicer partial applications
|
|
Fix (partial) #5502
|
|
We remove some dead aliases and add some documentation to the
interface file.
|
|
After #12504 , we can encapsulate and consolidate low-level state
logic in `Vernacstate`, removing `States` which is now a stub.
There is hope to clean up some stuff regarding the handling of
low-level proof state, by moving both `Evarutil.meta_counter` and
`Evd.evar_counter_summary` into the proof state itself [obligations
state is taken care in #11836] , but this will take some time.
|
|
Now that the interface has mostly stabilized, we move code around to
respect internal dependency order.
This will allow us to start sharing more code in the 4 principal
cases, and also paves the way for the full merging of obligations and
the removal of the Proof_ending type in favor of stronger type
abstraction.
|
|
We unify information about constants so it is shared among all the
paths [interactive, NI, obligations].
IMHO the current setup looks pretty good, with information split into
a per-constant record `CInfo.t` and variables affecting mutual
definitions at once, which live in `Info.t`.
Main information outside our `Info` record is `opaque`, which is
provided at different moments in several cases.
There are a few nits regarding interactive proofs, which will go away
in the next commits.
|
|
|
|
At this point the record in lemmas was just a stub; next commit will
stop exposing the internals of mutual information, and pave the way
for the refactoring of `Info.t` handling in the Declare interface.
|
|
|
|
After the previous commit, we don't need to chain a dummy future
anymore.
|
|
In the current proof save path, the kernel can raise an exception when
checking a proof wrapped into a future.
However, in this case, the future itself will "fix" the produced
exception, with the mandatory handler set at the future's creation
time.
Thus, there is no need for the declare layer to mess with exceptions
anymore, if my logic is correct. Note that the `fix_exn` parameter to
the `Declare` API was not used anymore.
This undoes 77cf18eb844b45776b2ec67be9f71e8dd4ca002c which comes from
pre-github times, so unfortunately I didn't have access to the
discussion.
We need to be careful in `perform_buildp` as to catch the `Qed` error
and properly notify the STM about it with `State.exn_on`; this was
previously done by the declare layer using a hack [grabbing internal
state of the future, that the future itself was not using as it was
already forced], but we now do it in the caller in a more principled
way.
This has been tested in the case that tactics succeed but Qed fails
asynchronously.
|
|
We complete some arduous refactoring in order to bring all the
internals and code of constant / proof saving into the same module.
In particular, this PR moves the remaining parts of proof saving from
`Lemmas` to `Declare`.
The reduction in exposed internals is considerable; in particular, we
remove the export of the internals of `proof_entry` and `proof_object`
[used in delayed proofs], which will allow us to start to address many
issues with the current setup, such as #10363 .
There are still some TODOs, that will be addressed in subsequent PRs:
- Remove `declare_constant` in favor of higher-level APIs
- Then, remove access to `proof_entry` entirely
- Refactor current very verbose handling of proof info.
- Remove compat modules / API.
- Rework handling of delayed proofs [this may be hard due to state and the STM]
- Reify Hook API for the case where it acts as a continuation [that is to say, declaring constants from the Hook]
List of remaining offenders for `proof_entry` / `declare_constant` in
the codebase:
- File "vernac/comHints.ml"
- File "vernac/indschemes.ml"
- File "vernac/comProgramFixpoint.ml"
- File "vernac/comAssumption.ml"
- File "vernac/record.ml"
- File "plugins/ltac/leminv.ml"
- File "plugins/setoid_ring/newring.ml"
- File "plugins/funind/recdef.ml"
- File "plugins/funind/gen_principle.ml"
|
|
This is needed as a first step to refactor and unify the obligation
save path and state; in particular `Equations` is a heavy user of
Hooks to modify obligations state, thus in order to make the hook
aware of this we need to place the obligation state before the hook.
As a good side-effect, `inline_private_constants` and `Hook.call` are
not exported from `Declare` anymore.
|
|
In our quest to unify all the declaration paths, an important step
is to account for the state pertaining to `Program` declarations.
Whereas regular proofs keep are kept in a stack-like structure;
obligations for constants defined by `Program` are stored in a global
map which is manipulated by almost regular open/close proof primitives.
This PR is in preparation for the switch to a purely functional state
in #11836 ; the full switch requires deeper changes so it is helpful
to have this PR preparing most of the structure.
Most of the PR is routine; only remarkable change is that the hook for
admitted obligations is now called explicitly in `finish_admitted` as
it had to learn about the different types of proof_endings. Before,
obligations set it in `start_lemma` but only used in the `Admitted`
path.
|
|
|
|
It was only used at one point in the STM, and its localization was suprising
to say the least. Furthermore it was relying on legacy code. Instead we hardcode
it in the STM.
|
|
This still needs API cleanup but we defer it to the moment we are
ready to make the internals private.
|
|
If we remove all the legacy proof engine stuff, that would remove the
need for the view on proof almost entirely.
|
|
We place creation and saving of interactive proofs in the same module;
this will allow to make `proof_entry` private, improving invariants
and control over clients, and to reduce the API [for example next
commit will move abstract declaration into this module, removing the
exported ad-hoc `build_constant_by_tactic`]
Next step will be to unify all the common code in the interactive /
non-interactive case; but we need to tweak the handling of obligations
first.
|
|
Following an observation by Enrico Tassi, we remove the `opaque`
parameter from `close_future_proof`, it should never be called with
transparent constants.
We will enforce this thru typing at the proof layer soon.
|
|
This is a small refactoring as these two functions behave very
differently and the invariants are quite different, in fact regular
`return_proof` should not be exported but be part of close proof, but
there is small use in the STM still.
|
|
We make the types of the delayed / non-delayed declaration path
different, as the latter is just creating futures that are forced
right away.
TTBOMK the new code should behave identically w.r.t. old one, modulo
the equation `Future.(force (from_val x)) = x`.
There are some questions as what the code is doing, but in this PR
I've opted to keep the code 100% faithful to the original one, unless
I did a mistake.
|
|
We make internal types `private` as an step towards the unification of
the save path with the rest of the system.
In particular, this is allow us to guarantee invariants
w.r.t. external users as the large majority of fields are always
constant.
This will also enable at some point a common creation of proof entry
with the rest of the system.
|
|
Add headers to a few files which were missing them.
|
|
Reviewed-by: Matafou
Reviewed-by: ppedrot
|
|
|
|
This PR refactors the handling of ML loadpaths to get it closer to
what (as of 2020) the standard OCaml toolchain (ocamlfind, dune) does.
This is motivated as I am leaning toward letting the standard OCaml
machinery handle OCaml includes; this has several benefits [for
example plugins become regular OCaml libs] It will also help in
improving dependency handling in plugin dynload.
The main change is that "recursive" ML loadpaths are no longer
supported, so Coq's `-I` option becomes closer to OCaml's semantics.
We still allow `-Q` to extend the OCaml path recursively, but this may
become deprecated in the future if we decide to install the ML parts
of plugins in the standard OCaml location.
Due to this `Loadpath` still hooks into `Mltop`, but other than that
`.v` location handling is actually very close to become fully
independent of Coq [thus it can be used in other tools such coqdep,
the build system, etc...]
In terms of vernaculars the changes are:
- The `Add Rec ML Path` command has been removed,
- The `Add Loadpath "foo".` has been removed. We now require that the
form with the explicit prefix `Add Loadpath "foo" as Prefix.` is used.
We did modify `fake_ide` as not to add a directory with the empty
`Prefix`, which was not used. This exposed some bugs in the
implementation of the document model, which relied on having an
initial sentence; we have workarounded them just by adding a dummy one
in the two relevant cases.
|
|
We make the primitives for backtrace-enriched exceptions canonical in
the `Exninfo` module, deprecating all other aliases.
At some point dependencies between `CErrors` and `Exninfo` were a bit
complex, after recent clean-ups the roles seem much clearer so we can
have a single place for `iraise` and `capture`.
|
|
|
|
Fixes #11608.
This means -vos doesn't skip proofs for definitions that end with Qed
but don't include Proof and rely on a Set Default Proof Using. However,
this fixes the bug where this pattern would instead hang, due to #11564.
|
|
Raising inside exception printers is quite tricky as the order of
registration for printers will indeed depend on the linking order.
We thus forbid this, and make our API closer to the upstream
`Printexn` by having printers return an option type.
|
|
Ack-by: gares
|
|
The standard use is to repeat the option keywords in lowercase, which
is basically useless.
En passant add doc entry for Dump Arith.
|
|
Treat a Set Default Proof Using call the same as having a syntactic
Proof using ... annotation. This ensures that proofs in sections are
skipped in -vos mode when there are enough annotations to determine the
type of the resulting theorem.
Fixes #11342.
|
|
|
|
This means return_proof is the only place where it can be produced.
We need to change the stm a bit as it keeps a pointer to a
[closed_proof_output] to join and to check if it failed, and it needs
a way to create a dummy in 1 case.
I'm not sure if this works correctly though, how to test?
We also inline the used bits of [return_proof ~allow_partial:true] in
[save_lemma_admitted] to get [Proof using] info. We could
alternatively expose the [closed_proof_output -> constr list]
projection. I think the code is easier to understand this way though,
as we don't have to read [return_proof] and figure out that the side
effect manipulation is ignored etc.
Note that the "this will warn" comment was outdated since
73daf37ccc7a44cd29c9b67405111756c75cb26a
|
|
|
