| Age | Commit message (Collapse) | Author |
|---|
|
Make proof-goto-point and proof-process-buffer prefix argument
aware. With argument, both commands temporarily switch off
proof-omit-proofs-option, such that all proofs are completely
processed for one particular invocation.
|
|
This commit adds a feature to recognize complete opaque proofs in
the asserted region and to replace them with an admitted proof.
This can drastically improve the processing time for the asserted
region at the cost of not checking the omitted proofs. Omitted
proofs are displayed slightly darker compared to other parts of
the locked region.
With this commit, the feature is supported for Coq for files in
which proofs are started with some form of Proof and ended with
either Qed, Defined, Admitted or Abort.
To enable, configure proof-omit-proofs-option or click
Proof General -> Quick Options -> Processing -> Omit Proofs.
|
|
Emacs occasionally loads Elisp files just to get more info (e.g. for C-h f),
so loading a file should "no effect". Fix the most obvious such effects:
the splash screen and the autotests by moving those effects into a function.
* coq/coq-autotest.el: Make it declarative. Use lexical-binding.
(coq-autotest): New function holding the code that used to be at top-level.
* generic/proof.el: Use lexical-binding.
Don't call proof-splash-message just because we're being loaded.
* generic/proof-script.el: Use lexical-scoping; fix all warnings.
(pg-show-all-portions): Simplify the code with a closure.
(proof-activate-scripting): Declare activated-interactively as dyn-scoped.
(proof--splash-done): New var.
(proof-mode): Call proof-splash-message upon first use.
* generic/proof-splash.el (proof-splash-message): Don't check
byte-compile-current-file now that we're only called when the mode
is activated.
* acl2/acl2.el (auto-mode-alist): Use `add-to-list` and \'.
* coq/coq-db.el (coq-build-menu-from-db-internal): Avoid silly O(N²).
* coq/coq-seq-compile.el:
* coq/coq-par-test.el:
* coq/coq-par-compile.el: Fix leftover uses of CL's `assert`.
* generic/proof-utils.el:
* generic/pg-movie.el:
* etc/testsuite/pg-test.el:
* coq/coq-syntax.el: Fix leftover uses of CL's `incf`.
* generic/pg-autotest.el: Fix leftover uses of CL's `decf`.
* obsolete/plastic/plastic.el (plastic-preprocessing): Fix leftover use
of CL's `loop`.
* generic/pg-user.el (proof-add-completions): Do nothing if no
proof-assistant is set yet (i.e. during byte-compilation).
(byte-compile-current-file): No need to test this any more.
* generic/proof-syntax.el (proof-regexp-alt-list): Use mapconcat.
Remove unnecessary "\\(?:...\\)".
(proof-regexp-alt): Redefine in terms of proof-regexp-alt-list.
|
|
|
|
|
|
The PG Makefile does ensure (using --batch) that noninteractive is non-nil while
compiling, but package.el doesn't.
|
|
|
|
|
|
|
|
'horizontal 'vertical 'smart policy.
|
|
immediately when opening a file.
|
|
|
|
|
|
|
|
|
|
who read or edit by waving mouse at text
|
|
Untested, see http://proofgeneral.inf.ed.ac.uk/trac/ticket/308
|
|
The following points are implemented already:
- recompile either via an external command (make) or let
ProofGeneral handle everything internally
- complete dependency tracking and recompilation for coq files in
internal mode
- support for extending the LoadPath: does almost work, even if
specified file-locally
- move back to clean state if recompilation fails
There are the following known problems:
- coq-load-path extensions are not retracted
- fails on partially qualified library names
|
|
|
|
prover specific.
|
|
(i.e., automatic preview of next command)
|
|
feature)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
