| Age | Commit message (Collapse) | Author |
|---|
|
"apply" unification.
Assuming w_unify_0 is not eventually abandoned, it remains to merge
unify_with_eta into unify_0 (what unify_with_eta does and that unify_0
does not do is to select of two instances of the same meta the one
with less lambda's; it is unclear whether this is useful heuristic).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@14091 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Some toplevel commands (for instance the experimental bullets) are
composed of several atomic commands, the failure of one must imply
the failure of the whole toplevel command. This commit introduces
a system of transaction to that effect.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@14087 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@14086 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
backward compatibility.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@14022 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
This reverts commit 33434695615806a85cec88452c93ea69ffc0e719.
Conflicts:
kernel/term_typing.ml
test-suite/success/polymorphism.v
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13998 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13988 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
use a lemma name chosen by the caller (here tclABSTRACT)
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13972 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
The recent experiment with -dont-load-proofs in the stdlib showed that
this options isn't fully safe: some axioms were generated (Include ?
functor application ? This is still to be fully understood).
Instead, I've implemented an idea of Yann: only load opaque proofs when
we need them. This is almost as fast as -dont-load-proofs (on the stdlib,
we're now 15% faster than before instead of 20% faster with -dont-load-proofs),
but fully compatible with Coq standard behavior.
Technically, the const_body field of Declarations.constant_body now regroup
const_body + const_opaque + const_inline in a ternary type. It is now either:
- Undef : an axiom or parameter, with an inline info
- Def : a transparent definition, with a constr_substituted
- OpaqueDef : an opaque definition, with a lazy constr_substitued
Accessing the lazy constr of an OpaqueDef might trigger the read on disk of
the final section of a .vo, where opaque proofs are located.
Some functions (body_of_constant, is_opaque, constant_has_body) emulate
the behavior of the old fields. The rest of Coq (including the checker)
has been adapted accordingly, either via direct access to the new const_body
or via these new functions. Many places look nicer now (ok, subjective notion).
There are now three options: -lazy-load-proofs (default), -force-load-proofs
(earlier semantics), -dont-load-proofs. Note that -outputstate now implies
-force-load-proofs (otherwise the marshaling fails on some delayed lazy).
On the way, I fixed what looked like a bug : a module type
(T with Definition x := c) was accepted even when x in T was opaque.
I also tried to clarify Subtyping.check_constant.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13952 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Note: even if this new tactical can be quite handy during the development phase,
(for instance to bound the time allocated to some search tactics), please be aware
of its main drawback: with it, scripts are no longer machine-independant, something
that works on a quick machine may fail on a slow one. The converse is even possible
if you combine this "timeout" with other tactic combinators. We strongly advise to
not leave any "timeout" in the final version of a development.
In addition, this feature won't probably work on native win32, since Unix.alarm
isn't implemented.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13917 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
conversion when checking types of instanciations while having
restricted delta reduction for unification itself. This
makes auto/eauto... backward compatible.
- Change semantics of [Instance foo : C a.] to _not_ search
for an instance of [C a] automatically and potentially slow
down interaction, except for trivial classes with no fields.
Use [C a := _.] or [C a := {}] to search for an instance of
the class or for every field.
- Correct treatment of transparency information for classes
declared in sections.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13908 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
in the 8.3 patch.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13903 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
unification failure messages (it is not fully usable and was not
intended to be committed now, sorry for the noise).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13895 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
error messages. The architecture of unification error handling
changed, not helped by ocaml for checking that every exceptions is
correctly caught. Report or fix if you find a regression.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13893 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Before this patch, hints such as "Hint Resolve -> a" in success/Hints.v
were erroneously considered "eauto-only". We try to clarify the big
boolean expression via "if", and for the moment we remove
the detection of "nonlinearity" via duplicated_metas : on the example,
some nonlinearity was found for strange reason (beta expansion ?), and
after some discussion with Hugo, it is unclear whether this nonlinearity
stuff is useful at all. The next coqbench might provide some answer
to this question, we'll see
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13850 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
conversion.
- Fix trans_fconv* to use evars correctly.
- Normalize the goal with respect to evars before rewriting in
[rewrite], allowing to see instanciations from other subgoals.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13844 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Everything seems to work fine in CoqIDE (except escape/return and the daimon which are not entirely ported).
However, there is some problem causing proof general to fail when using goto or evaluate buffer (evaluate next phrase works fine though), as well as coqc.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13817 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
new proof engine. Correct treatment of the evar set: the tactic
incrementally extends (and potentially refines) the existing sigma and
the internally generated typeclasses constraints are removed from it at
the end as they are always solved. This avoids tricky and costly
evar_map manipulations.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13812 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
According to B. Gregoire, this stuff is obsolete. Fine control
on when to launch the VM in conversion problems is now provided
by VMcast. We were already almost never boxing definitions anymore
in stdlib files.
"(Un)Boxed Definition foo" will now trigger a parsing error,
same with Fixpoint. The option "(Un)Set Boxed Definitions"
aren't there anymore, but tolerated (as no-ops), since unknown
options raise a warning instead of an error by default.
Some more cleaning could be done in the vm.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13806 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
perl -pi -e 's/(\W|_)raw((?:sort|_prop|terms?|_branch|_red_flag|pat
tern|_constr_of|_of_pat)(?:\W|_))/\1glob_\2/g;s/glob__/glob_/g;s/(\
W)R((?:Prop|Type|Fix|CoFix|StructRec|WfRec|MeasureRec)\W)/\1G\2/g;s
/glob_terms?/glob_constr/g' **/*.ml*
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13756 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13744 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
There was a discrepancy of the notions "raw" and "globalized" between
constrs and tactics, and some confusion of the notions in
e.g. genarg.mli (see all globwit_* there). This commit is a first step
towards unification of terminology between constrs and
tactics. Changes in module names will be done separately.
In extraargs.ml4, the "ARGUMENT EXTEND raw" and related stuff, even
affected by this change, has not been touched and highlights another
confusion in "ARGUMENT EXTEND" in general that will be addressed
later.
The funind plugin doesn't respect the same naming conventions as the
rest, so leave some "raw" there for now... they will be addressed
later.
This big commit has been generated with the following command (wrapped
here, but should be on a *single* line):
perl -pi -e 's/(\W(?:|pp|pr_l)|_)raw((?:constrs?|type|vars|_binder|
_context|decl|_decompose|_compose|_make)(?:\W|_))/\1glob_\2/g;s/glo
b__/glob_/g;s/prraw/prglob/g;s/(\W)R((?:Ref|Var|Evar|PatVar|App|Lam
bda|Prod|LetIn|Cases|LetTuple|If|Rec|Sort|Hole|Cast|Dynamic)\W)/\1G
\2/g' `git ls-files|grep -v dev/doc/changes.txt`
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13743 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Apparently, it seems that clenv.evd is either created
from dummy_goal (in (e)auto) or from a copy of gls (in
class_tactics). I've checked experimentally by some assert
that on the stdlib the defined part of clenv.evd is always
included in gls. I hence propose to simplify this function
connect_clenv.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13718 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13710 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
placed by vm_cast_no_check
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13703 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
before the message is delivered to the user. Should avoid useless
computation in heavily backtracking tactics (auto, try, ...).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13628 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Patch from Adam Megacz.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13577 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
It looks obsolete, probably as a consequence of the new proof engine.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13504 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
In particular, the unused lib/tlm.ml and lib/gset.ml are removed
In addition, to simplify code, Libobject.record_object returning only the
('a->obj) function, which is enough almost all the time.
Use Libobject.record_object_full if you really need also the (obj->'a).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13460 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
governed in the latter case by a flag since (useful e.g. for setoid
rewriting which otherwise loops as it is implemented).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13443 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
user either gives all missing arguments not dependent in the concl or
all missing arguments not *recursively* dependent in the concl (as
introduced by commit 13367). In practice, this means that "apply
f_equal with A" remains allowed even though the new, recursive,
analysis detects that all arguments of f_equal are inferable,
including the first type argument (which is inferable from the
knowledge of the function).
Sized the opportunity to better explain the behavior of clenv_dependent.
Also made minor code simplification.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13426 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
dependent arguments. It breaks compatibility as some [apply with]
clauses are not necessary anymore. Typically when applying [f_equal],
the domain type of the function can be infered even if it does not
appear directly in the conclusion of the goal. Fixes bug #2154.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13367 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13323 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
engine.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13300 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Applied it to fix mli file headers.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13176 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13168 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
enough)
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13126 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
generalize the interface of Clenv to be able to use the existing
treatment of bindings. Clenv functions did not use goals conclusions
but insisted on getting goals anyway (which is even more problematic as
goals appear in evar maps now).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13102 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
Definition's is not so painless. It seems to however generally provide
"nicer" scripts so let us keep it and update the contribs and
test-suite accordingly.
Also enforced that the actual introduced names to be exactly as given
in the statements.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13097 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13085 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
- Instances found by matching.ml now collect the set of bound
variables they possibly depend on in the pattern (see type
Pattern.extended_patvar_map); the variables names are canonically
ordered so that non-linear matching takes actual names into account.
- Removed typing of matching constr instances in advance (in
tacinterp.ml) and did it only at use time (in pretyping.ml). Drawback
is that we may have to re-type several times the same term but it is
necessary for considering terms with locally bound variables of which
we do not keep the type (and if even we had kept the type, we would have
to adjust the indices to the actual context the term occurs).
- A bit of documentation of pattern.mli, matching.mli and pretyping.mli.
- Incidentally add env while printing idtac messages. It seems more correct
and I hope I did not break some intended existing behavior.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13080 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13037 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
The choice between camlp4/5 is done during configure with flags
-usecamlp5 (default for the moment) vs. -usecamlp4.
Currently, to have a full camlp4 compatibility, you need to change
all "EXTEND" and "GEXTEND Gram" into "EXTEND Gram", and change "EOI"
into "`EOI" in grammar entries. I've a sed script that does that
(actually the converse), but I prefer to re-think it and check a few
things before branching this sed into the build mechanism.
lib/compat.ml4 is heavily used to hide incompatibilities between camlp4/5
and try to propose a common interface (cf LexerSig / GrammarSig).
A few incompatible quotations have been turned into underlying code
manually, in order to make the IFDEF CAMLP5 THEN ... ELSE ... END
parsable by both camlp4 and 5. See in particular the fate of
<:str_item< declare ... end >>
Stdpp isn't used anymore, but rather Ploc (hidden behind local module Loc).
This forces to use camlp5 > 5.01.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13019 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
evar_map into a map for defined evars and a map for undefined evars.
Even before Spiwack's new proof engine, some Evd.fold were very
costly, e.g. in check_evars or progress_evar_map. With the new proof
engine, undefined evars traversals are apparently even more common (at
least, it improves significantly the complexity of some calls to omega
in JordanCurveTheorem - a new factor 5-7 after the factor 5-6 obtained
by removal of evar_merge in clenv_fchain in commit 13007, arriving to
figures comparable to the 8.3 ones).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13011 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13008 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
is an important cause of inefficiency when the number of evars is large.
It is wrong in the sense that it assumes the two sigma given to
clenv_fchain to be disjoint. This is true for metas, but not for
evars. It used to be useful in some respect when clenv bindings were
represented by open_constr, each of them having private evars (see
r10151 which introduced this evar_merge), but determining what evars
were private which and were shared is hopeless. Since the removal of
private sigmas in r12603, evars in clenv bindings are assumed to be
extended monotonically and clenv_fchain should only have to take the
most recent evars - assumed to be the first argument - instead of a
union.
The function clenv_fchain remains anyway fragile since it is
asymmetric. More should be done to clean it up so that it receives
only one - unambiguous - evars argument, what would mean actually,
receiving only one clausenv and the second argument being just a
template pair (t:T).
The call to evar_merge was source of inefficiency. On some calls to
omega in contrib JordanCurveTheorem, removing it reduces the execution
time by a factor larger than 5 (from 400s down to 70s on my Core 2 Duo).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@13007 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
- ide doesn't crash anymore at any backtrack
- I don't see if vernacentries.ml did the same assumption so I didn't
change anything. (The only other use of resume_proof)
- ide still raises "NoCurrentProof" when you type a bad keyword such as
"Priint"... But at least, this on is catch somewhere !
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@12993 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@12976 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
- Many of them were broken, some of them after Pierre B's rework
of mli for ocamldoc, but not only (many bad annotation, many files
with no svn property about Id, etc)
- Useless for those of us that work with git-svn (and a fortiori
in a forthcoming git-only setting)
- Even in svn, they seem to be of little interest
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@12972 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
dev/ocamlweb-doc has been erased. I hope no one still use the
"new-parse" it generate.
In dev/,
make html will generate in dev/html/ "clickable version of mlis". (as
the caml standard library)
make coq.pdf will generate nearly the same awfull stuff that coq.ps was.
make {kernel,lib,parsing,..}.{dot,png} will do the dependancy graph of
the given directory.
ocamldoc comment syntax is here :
http://caml.inria.fr/pub/docs/manual-ocaml/manual029.html
The possibility to put graphs in pdf/html seems to be lost.
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@12969 85f007b7-540e-0410-9357-904b9bb8a0f7
|
|
This is a fairly large commit (around 140 files and 7000 lines of code
impacted), it will cause some troubles for sure (I've listed the know
regressions below, there is bound to be more).
At this state of developpement it brings few features to the user, as
the old tactics were
ported with no change. Changes are on the side of the developer mostly.
Here comes a list of the major changes. I will stay brief, but the code
is hopefully well documented so that it is reasonably easy to infer the
details from it.
Feature developer-side:
* Primitives for a "real" refine tactic (generating a goal for each
evar).
* Abstract type of tactics, goals and proofs
* Tactics can act on several goals (formally all the focused goals). An
interesting consequence of this is that the tactical (. ; [ . | ... ])
can be separated in two
tacticals (. ; .) and ( [ . | ... ] ) (although there is a conflict for
this particular syntax). We can also imagine a tactic to reorder the
goals.
* Possibility for a tactic to pass a value to following tactics (a
typical example is
an intro function which tells the following tactics which name it
introduced).
* backtracking primitives for tactics (it is now possible to implement a
tactical '+'
with (a+b);c equivalent to (a;c+b;c) (itself equivalent to
(a;c||b;c)). This is a valuable
tool to implement tactics like "auto" without nowing of the
implementation of tactics.
* A notion of proof modes, which allows to dynamically change the parser
for tactics. It is controlled at user level with the keywords Set
Default Proof Mode (this is the proof mode which is loaded at the start
of each proof) and Proof Mode (switches the proof mode of the current
proof) to control them.
* A new primitive Evd.fold_undefined which operates like an Evd.fold,
except it only goes through the evars whose body is Evar_empty. This is
a common operation throughout the code,
some of the fold-and-test-if-empty occurences have been replaced by
fold_undefined. For now,
it is only implemented as a fold-and-test, but we expect to have some
optimisations coming some day, as there can be a lot of evars in an
evar_map with this new implementation (I've observed a couple of
thousands), whereas there are rarely more than a dozen undefined ones.
Folding being a linear operation, this might result in a significant
speed-up.
* The declarative mode has been moved into the plugins. This is made
possible by the proof mode feature. I tried to document it so that it
can serve as a tutorial for a tactic mode plugin.
Features user-side:
* Unfocus does not go back to the root of the proof if several Focus-s
have been performed.
It only goes back to the point where it was last focused.
* experimental (non-documented) support of keywords
BeginSubproof/EndSubproof:
BeginSubproof focuses on first goal, one can unfocus only with
EndSubproof, and only
if the proof is completed for that goal.
* experimental (non-documented) support for bullets ('+', '-' and '*')
they act as hierarchical BeginSubproof/EndSubproof:
First time one uses '+' (for instance) it focuses on first goal, when
the subproof is
completed, one can use '+' again which unfocuses and focuses on next
first goal.
Meanwhile, one cas use '*' (for instance) to focus more deeply.
Known regressions:
* The xml plugin had some functions related to proof trees. As the
structure of proof changed significantly, they do not work anymore.
* I do not know how to implement info or show script in this new engine.
Actually I don't even know what they were suppose to actually mean in
earlier versions either. I wager they would require some calm thinking
before going back to work.
* Declarative mode not entirely working (in particular proofs by
induction need to be restored).
* A bug in the inversion tactic (observed in some contributions)
* A bug in Program (observed in some contributions)
* Minor change in the 'old' type of tactics causing some contributions
to fail.
* Compilation time takes about 10-15% longer for unknown reasons (I
suspect it might be linked to the fact that I don't perform any
reduction at QED-s, and also to some linear operations on evar_map-s
(see Evd.fold_undefined above)).
git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@12961 85f007b7-540e-0410-9357-904b9bb8a0f7
|
