| Age | Commit message (Collapse) | Author |
|---|
|
off by default
|
|
We use heaps instead of continuously adding elements to an ordered list,
which was quadratic in the worst case.
As a byproduct, this solves bug #5359, which was due to a stack overflow on
big lists.
|
|
Several cleanups were performed.
1. Removal of dead code lurking around.
2. Removal of global variables used to pass arguments to functions, as well as
unnecessary mutable state here and there. We rely on state-passing and
encapsulated mutable state.
3. Removal of crazy reference manipulation and its replacement with proper list
handling, as well as cleaning up the source and taking advantage of invariants.
This should solve algorithmic limitations of the previous code.
4. Opacification of some structures to have a clearer idea of the code
requirements.
5. Cleaning of debug printing functions. We thunk the computation of the
debugging data, whose computation can be costly for no reason, and we rely
on Feedback-based interaction instead of Printf-debugging.
|
|
- Supporting let-ins in tactic "fix", and hence in interactive
Fixpoint and mutual theorems.
- Documenting more precisely the meaning of n in tactic "fix id n".
- Fixing computation of recursive index at interpretation time in the
presence of let-ins.
|
|
in the presence of let-ins.
|
|
|
|
Also documenting how the implicit arguments by position are computed.
|
|
|
|
There were actually no syntax for it, and I'm still unsure what good
syntax to give to it, even more that it would be useful to have one.
|
|
Instead of calling the whole reduction machirery, we check before reducing that
a term is an applied atom, i.e. inductive, constructor, evar or meta. In that
case, the abstract machine acts as the identity but needs to destruct and
reconstruct the whole term, which can be very costly.
This fixes part of bug #5421: vm_compute is very slow at doing nothing, where
recomputation of the type of a big inductive was incredibly expensive.
|
|
This was assuming dependencies occurring in configurations of
the form x:A, y:B x, z:C x y |- match x, y, z with ... end".
But still work to do for better management of dependencies in general...
|
|
8e07227c5853de78eaed4577eefe908fb84507c0 introduced an
incorrect duplicate of `position_error_tag_at_sentence`, which
sets the end of the underlining position starting at the end of the
sentence, whereas the location in the feedback refers to the
beginning, thus it highlights more text than it should.
This was missed in 8.6 as it seems that the code was not called.
We undo the duplication and fix the bug.
|
|
|
|
The evarmap used by the heuristic could contain resolved evars, which could
lead to a failure of backtracking in the EConstr branch. This is experimental
and may be to costly.
|
|
`process_error_hook` seems unnecesary, we just call the proper error
interpretation.
|
|
`tactic_being_run_hook` was used for the "xml" pluging but I am not
sure we have a sensible use case here.
|
|
|
|
This fixes the bug in `Drop` reported by @mattam82: after performing a
`Drop`, the feeder was lost and no further message from Coq was
printed.
|
|
|
|
This allows e.g. to use the record notations even when there are
defined fields.
A priori fixed also missing parameters when interpreting primitive
tokens.
|
|
|
|
|
|
comparisons.
|
|
|
|
|
|
This was making the miniopt target fail.
|
|
|
|
|
|
|
|
This removes some remaining support for camlp4 in the infrastructure
and documents the change.
|
|
We remove the camlp4 compatibility layer, and try to clean up
most structures. `parsing/compat` is gone.
We added some documentation to the lexer/parser interfaces that are
often obscured by module includes.
|
|
printer.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Getting a key only needs to observe the root of a term. This hotspot was
observed in HoTT.
|
|
|
|
|
|
This fixes a few clear bugs, but the STM code handling Time, Fail and
Redirect before par: still needs to be rewritten. It does not implement
the same semantics as the vernac interpreter for Fail Fail [c] and
ignores Redirect.
This commit was already reviewed with Enrico and tested on Travis.
|
|
|
|
|
|
Was needed to be done for a while.
|
|
We solve https://coq.inria.fr/bugs/show_bug.cgi?id=4789 by printing
all the errors from the feedback handler, even in the case of coqtop.
All error display is handled by a single, uniform path.
There may be some minor discrepancies with 8.6 as we are uniform now
whereas 8.6 tended to print errors in several ways, but our behavior
is a subset of the 8.6 behavior.
We had to make a choice for `-emacs` error output, which used to vary
too. We have chosen to display error messages as:
```
(location info) option \n
(program caret) option \n
MARKER[254]Error: msgMARKER[255]
```
This commit also fixes:
- https://coq.inria.fr/bugs/show_bug.cgi?id=5418
- https://coq.inria.fr/bugs/show_bug.cgi?id=5429
|
|
It was not necessary to normalize a term just to check whether it was a
global reference. The hotspot appeared in mathcomp.
|
|
Note: Apparently not easy to make a test file as the error is raised
in "G_vernac.fresh_var" at parsing time (not captured by Fail).
|
|
|
