leanprover / lean

Lean Theorem Prover

This repository has been frozen. Lean 4 is now the official release.

About

• Homepage
• Theorem Proving in Lean
• FAQ

Installation

Stable and nightly binary releases of Lean are available on the homepage. For building Lean from source, see the build instructions.

lean's Issues

lean server crashes on OSX

On OSX, lean server crashes immediately when it opens a file. The following is an example input.

VISIT /Users/soonhok/work/lean/library/standard/logic/axioms/hilbert.lean

The following is the traces from valgrind:

$ valgrind ~/work/lean/bin/lean --server < input                                                                                           [22:27:42 on 14-08-15] ruby-1.9.3-p327
==31497== Memcheck, a memory error detector
==31497== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
==31497== Using Valgrind-3.9.0 and LibVEX; rerun with -h for copyright info
==31497== Command: /Users/soonhok/work/lean/bin/lean --server
==31497==
==31497== WARNING: Support on MacOS 10.8/10.9 is experimental and mostly broken.
==31497== WARNING: Expect incorrect results, assertions and crashes.
==31497== WARNING: In particular, Memcheck on 32-bit programs will fail to
==31497== WARNING: detect any errors associated with heap-allocated data.
==31497==
--31497-- /Users/soonhok/work/lean/bin/lean:
--31497-- dSYM directory is missing; consider using --dsymutil=yes
!!!Interrupted!!!
==31497== Thread 2:
==31497== Invalid read of size 8
==31497==    at 0x1000EC3E3: thread_specific_ptr<lean::memory_pool<48u> >::~thread_specific_ptr() (in /Users/soonhok/work/lean/bin/lean)
==31497==    by 0xFD485: (anonymous namespace)::run(void*) (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libstdc++.6.dylib)
==31497==    by 0x620478: _pthread_exit (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x61F8A3: _pthread_body (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x61F729: _pthread_start (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x623FC8: thread_start (in /usr/lib/system/libsystem_pthread.dylib)
==31497==  Address 0x100f6c998 is 8 bytes inside a block of size 16 free'd
==31497==    at 0x722F: free (in /usr/local/Cellar/valgrind/3.9.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==31497==    by 0x2CB802: emutls_destroy (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libgcc_s.1.dylib)
==31497==    by 0x101: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0xB0080FFF: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==
==31497== Invalid read of size 8
==31497==    at 0x1000EC3EF: thread_specific_ptr<lean::memory_pool<48u> >::~thread_specific_ptr() (in /Users/soonhok/work/lean/bin/lean)
==31497==    by 0xFD485: (anonymous namespace)::run(void*) (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libstdc++.6.dylib)
==31497==    by 0x620478: _pthread_exit (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x61F8A3: _pthread_body (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x61F729: _pthread_start (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x623FC8: thread_start (in /usr/lib/system/libsystem_pthread.dylib)
==31497==  Address 0x100f6c998 is 8 bytes inside a block of size 16 free'd
==31497==    at 0x722F: free (in /usr/local/Cellar/valgrind/3.9.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==31497==    by 0x2CB802: emutls_destroy (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libgcc_s.1.dylib)
==31497==    by 0x101: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0xB0080FFF: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==
==31497== Invalid write of size 8
==31497==    at 0x1000EC40B: thread_specific_ptr<lean::memory_pool<48u> >::~thread_specific_ptr() (in /Users/soonhok/work/lean/bin/lean)
==31497==    by 0xFD485: (anonymous namespace)::run(void*) (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libstdc++.6.dylib)
==31497==    by 0x620478: _pthread_exit (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x61F8A3: _pthread_body (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x61F729: _pthread_start (in /usr/lib/system/libsystem_pthread.dylib)
==31497==    by 0x623FC8: thread_start (in /usr/lib/system/libsystem_pthread.dylib)
==31497==  Address 0x100f6c998 is 8 bytes inside a block of size 16 free'd
==31497==    at 0x722F: free (in /usr/local/Cellar/valgrind/3.9.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==31497==    by 0x2CB802: emutls_destroy (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libgcc_s.1.dylib)
==31497==    by 0x101: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0xB0080FFF: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==
==31497== Invalid read of size 8
==31497==    at 0x10013265D: thread_specific_ptr<lean::memory_pool<32u> >::~thread_specific_ptr() (in /Users/soonhok/work/lean/bin/lean)
==31497==    by 0xFD485: (anonymous namespace)::run(void*) (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libstdc++.6.dylib)
==31497==    by 0x101: ???
==31497==    by 0x102: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==  Address 0x100f63718 is 8 bytes inside a block of size 16 free'd
==31497==    at 0x722F: free (in /usr/local/Cellar/valgrind/3.9.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==31497==    by 0x2CB802: emutls_destroy (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libgcc_s.1.dylib)
==31497==    by 0x101: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0xB0080FFF: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==
==31497== Invalid read of size 8
==31497==    at 0x100132669: thread_specific_ptr<lean::memory_pool<32u> >::~thread_specific_ptr() (in /Users/soonhok/work/lean/bin/lean)
==31497==    by 0xFD485: (anonymous namespace)::run(void*) (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libstdc++.6.dylib)
==31497==    by 0x101: ???
==31497==    by 0x102: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==  Address 0x100f63718 is 8 bytes inside a block of size 16 free'd
==31497==    at 0x722F: free (in /usr/local/Cellar/valgrind/3.9.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==31497==    by 0x2CB802: emutls_destroy (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libgcc_s.1.dylib)
==31497==    by 0x101: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0xB0080FFF: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==
==31497== Invalid write of size 8
==31497==    at 0x100132685: thread_specific_ptr<lean::memory_pool<32u> >::~thread_specific_ptr() (in /Users/soonhok/work/lean/bin/lean)
==31497==    by 0xFD485: (anonymous namespace)::run(void*) (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libstdc++.6.dylib)
==31497==    by 0x101: ???
==31497==    by 0x102: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==  Address 0x100f63718 is 8 bytes inside a block of size 16 free'd
==31497==    at 0x722F: free (in /usr/local/Cellar/valgrind/3.9.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==31497==    by 0x2CB802: emutls_destroy (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libgcc_s.1.dylib)
==31497==    by 0x101: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0xB0080FFF: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==
==31497== Invalid read of size 8
==31497==    at 0x1001BB213: lean::list<lean::options>::~list() (in /Users/soonhok/work/lean/bin/lean)
==31497==    by 0xFD485: (anonymous namespace)::run(void*) (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libstdc++.6.dylib)
==31497==    by 0x101: ???
==31497==    by 0x102: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==  Address 0x100f574c8 is 8 bytes inside a block of size 16 free'd
==31497==    at 0x722F: free (in /usr/local/Cellar/valgrind/3.9.0/lib/valgrind/vgpreload_memcheck-amd64-darwin.so)
==31497==    by 0x2CB802: emutls_destroy (in /usr/local/Cellar/gcc/4.9.1/lib/gcc/x86_64-apple-darwin13.2.0/4.9.1/libgcc_s.1.dylib)
==31497==    by 0x101: ???
==31497==    by 0xB0080E9F: ???
==31497==    by 0xB0080FFF: ???
==31497==    by 0x62354E: _pthread_tsd_cleanup (in /usr/lib/system/libsystem_pthread.dylib)
==31497==
==31497==
==31497== HEAP SUMMARY:
==31497==     in use at exit: 154,708 bytes in 1,336 blocks
==31497==   total heap usage: 5,207 allocs, 3,871 frees, 625,546 bytes allocated
==31497==
==31497== LEAK SUMMARY:
==31497==    definitely lost: 0 bytes in 0 blocks
==31497==    indirectly lost: 0 bytes in 0 blocks
==31497==      possibly lost: 594 bytes in 30 blocks
==31497==    still reachable: 128,364 bytes in 931 blocks
==31497==         suppressed: 25,750 bytes in 375 blocks
==31497== Rerun with --leak-check=full to see details of leaked memory
==31497==
==31497== For counts of detected and suppressed errors, rerun with: -v
==31497== ERROR SUMMARY: 7 errors from 7 contexts (suppressed: 2003 from 107)

lean-mode: auto-complete doesn't work at the end of buffer

--flyinfo shows numeric representation of meta variables

FLYCHECK_BEGIN INFO
/home/soonhok/work/lean/library/standard/logic/axioms/hilbert.lean:23:68: type
Exists (λ (f : Π (3.1.14 : A), B 3.1.14), ∀ (x : A), P x (f x))
FLYCHECK_END

type info left-over

pre elaboration type information is not being deleted when elaboration succeeds

shorter commands, like "lean foo" and "lean foo -o"

output convention for type info

Can we change the output convention for type information?

• Option name: --typeinfo
• --typeinfo takes an optional argument --typeinfo <LINE_NO>: If LINE_NO is specified, lean only reports type information at the specified line LINE_NO, otherwise it reports all type information in a file.
• Message format: use | as a delimiter. Note that <TYPE_OF_ID> doesn't have a newline \nin it

TYPE|<FILENAME>|<LINE_NO>|<COL>|<TYPE_OF_ID>

Example:

TYPE|/home/soonhok/work/lean/library/standard/logic/connectives/basic.lean|154|69|(a → b) → (b → a) → and (a → b) (b → a)
TYPE|/home/soonhok/work/lean/library/standard/logic/connectives/basic.lean|154|79|a → b
TYPE|/home/soonhok/work/lean/library/standard/logic/connectives/basic.lean|154|82|b → a

lmake crash

lmake crashes if we execute it in any file at tests/lean/run. Here is the trace

Traceback (most recent call last):
File "../../../bin/lmake", line 187, in
sys.exit(main())
File "../../../bin/lmake", line 178, in main
directory = os.path.dirname(makefile)
File "/usr/lib/python2.7/posixpath.py", line 120, in dirname
i = p.rfind('/') + 1
AttributeError: 'bool' object has no attribute 'rfind'

add fake dependency to LEAN_VERSION fake file

The idea is to force the library to be recompiled whenever we update the executable/version

apply tactic

We need to re-implement it and cleanup.
We also need a higher-order unification procedure that is independent of the elaborator because it does too much. For example, we don't want to unfold definitions or apply beta-reduction.

emacs configuration

More emacs configuration could go in either the lean emacs mode or the flycheck mode. Two things that come to mind are:

(1) deleting trailing whitespace
(2) putting the error message window and compilation window to the right, by default
(3) setting the file window to 100 columns, by default

error message for application

With

definition test {A B C : Type} (f : A → B) (x : C) := f x

I suggest the error message

type mismatch at application
f x
term
x
has type
C
but is expected to have type
A

This changes the current order of A and C, and also the wording slightly.

lean server: caching and implicit arguments

open a file and enter the following:

import logic
theorem test (T : Type) (x : T) : x = x := refl _
theorem test2 (A : Type) (a : A) : a = a := test A a

Then make the first argument to test implicit:

theorem test {T : Type} (x : T) : x = x := refl _

Flycheck fails to report a error in test2 unless / until test2 is actually changed. With

#erase_cache test2

the behavior is corrected.

emacs: support -- NAY

default for inductive types

With

inductive ... : Type :=
...

we could allow the user to omit ": Type".

unambiguous coercions

theorem test [coercion] ... : A -> B

is currently flagged as ambiguous, but the colon makes it unambiguous.

SMT frontend

Parser for SMT 2.0

section - end matching

Let's consider

section foo
...
end foo

and

namespace foo
...
end foo

Should we require that all sections and namespaces be closed at the end of a file?

use cache in server mode

(bitbucket) email diff is disabled.

Recently, we had a problem in pushing to bitbucket in travis-ci, and disabled the feature. One side effect of it is that we don't get commit email notification with full diff of changes. For now, I enabled github's email notification feature. Unfortunately, github does not support email diff and has no plan to do it. I'll see how to fix the original bitbucket problem.

generate corrupted .olean file even if compilation fails

make for directory structure

import file - relative indexing

namespace management

Implement the following behavior for namespaces:

  namespace foo
    theorem bar : ...
  end foo

creates identifier foo.bar.

"import" and "namespace" are kept orthogonal (file structure has nothing to do with namespace structure).

The command

  open foo [as bar] [using / hiding / renaming]

does the following (conceptually):

(1) gathers all the theorems in foo, removing "foo"
(2) picks out the ones indicated by using / hiding
(3) renames according to renaming
(4) creates aliases, prepending "bar" if present

The command

  export foo [as bar] [using / hiding / renaming]

does the same thing but creates permanent abbreviations rather than transient aliases.

Question: should we have special syntax for opening / exporting a single theorem? E.g.

  open foo bar

or

  open foo (bar)

for

  open foo (using bar)

Inside a module, a theorem or definition can be marked as "private", which means that it is never opened or exported. It can also be marked as "protected", which means that it is never opened or exported unless the user explicitly asks for it.

Question: what should the syntax for that be? We can require that it is on an explicit "using" list. In that case, the user would have to say

  open nat
  open nat (using induction_on)

possibly with the shorthand above. Another option is to add another modifier, "including", which can be used in the absence of using. So

  open nat (including induction_on)

would have the same meaning as the two lines above.

Add options to flag an error or warning if identifies become overloaded, and also if notation conflicts.

A command

  theorem foo [in bar] ...

puts the theorem in bar (i.e. calls it bar.foo) but also uses it, i.e. creates an alias.

The command

  inductive foo : Type :=
  | bar : ...
  | baz : ...

creates constants foo, foo.mk, foo.rec, foo.bar, foo.baz, and similarly for "structure".

The construct

  section

  end

should be used to set parameters common to all theorems. Notations, rewrites, and so on are persistent.

The construct

  context

  end

should be used to set up a local context. Notations, rewrites, and so on are transient.

lmake error message

Making dependency file '/home/jeremy/projects/lean/library/standard/struc/wf.d' ...
error: file 'data/pair' not found in the LEAN_PATH
Making dependency file '/home/jeremy/projects/lean/library/standard/struc/binary.d' ...

generalize and intro tactics

The tactic language needs these (like Coq's).

add support for indexing: .ilean files should be generated automatically

problem with 'INSERT' command (lean-server)

I was testing the lean-show-type function, and found a problem with INSERT command.

First, open library/standard/data/list/basic.lean file and send a query about line 280:

VISIT /home/soonhok/work/lean/library/standard/data/list/basic.lean
INFO 280
/home/soonhok/work/lean/library/standard/data/list/basic.lean:17:0: warning: imported file uses 'sorry'
-- BEGININFO
-- TYPE|280|21
Type
-- ACK
-- TYPE|280|29
Type → Type
-- ACK
-- TYPE|280|34
Type
-- ACK
-- TYPE|280|42
Type
-- ACK
-- TYPE|280|47
Type
-- ACK
-- ENDINFO

Good, let's insert "-- comment" at line 280.

INSERT 280
-- comment

I expect we have the same information at line 281 (since the old line 280 is shifted to 281)

INFO 281
/home/soonhok/work/lean/library/standard/data/list/basic.lean:225:0: warning: imported file uses 'sorry'
-- BEGININFO
-- ENDINFO

There is nothing. To make sure, I send more queries:

INFO 280
-- BEGININFO
-- ENDINFO
INFO 279
-- BEGININFO
-- ENDINFO
INFO 282
-- BEGININFO
-- ENDINFO
INFO 283
-- BEGININFO
-- ENDINFO

Note that I added -- BEGININFO at the beginning of INFO messages.

ltags

ltags takes *.ilean files and generates TAGS file, possibliy supporting multiple tag systems such as etags and global.

type information for placeholders

add -- SYNTH

Thread-local storage unsupported for the current target: problem in compilation

Caro Leonardo,

Tentei compilar o Lean no meu Mac OS Mavericks, seguindo as instruções aqui: https://github.com/leodemoura/lean/blob/master/doc/make/cmake_make.md

cmake deu certo, mas make falhou. Segue abaixo a mensagem de erro:

release$ cmake -DCMAKE_BUILD_TYPE=RELEASE ../../src
-- Could NOT find TCMALLOC (missing:  TCMALLOC_INCLUDE_DIR TCMALLOC_LIBRARIES) 
*** WARNING: failed to find tcmalloc
*** The (optional) tcmalloc library is available at: https://code.google.com/p/gperftools
CMake Warning at CMakeLists.txt:125 (message):
  FAILED to find tcmalloc, using standard malloc.


-- Usable malloc_usable_size was not detected
-- Found malloc_size 
-- Found Lua: /usr/local/lib/liblua.dylib (found version "5.1.5")
-- lua_newstate works
-- Found PythonInterp: /Library/Frameworks/Python.framework/Versions/2.7/bin/python (found version "2.7.5") 
-- Configuring done
-- Generating done
-- Build files have been written to: /Users/Bruno/Dropbox/Code/Lean/build/release
release$ make
Scanning dependencies of target util
[  0%] Building CXX object util/CMakeFiles/util.dir/trace.cpp.o
[  1%] Building CXX object util/CMakeFiles/util.dir/debug.cpp.o
[  1%] Building CXX object util/CMakeFiles/util.dir/name.cpp.o
[  1%] Building CXX object util/CMakeFiles/util.dir/name_set.cpp.o
[  2%] Building CXX object util/CMakeFiles/util.dir/exception.cpp.o
/Users/Bruno/Dropbox/Code/Lean/src/util/exception.cpp:29:16: error: 
      thread-local storage is unsupported for the current target
        static LEAN_THREAD_LOCAL std::string buffer;
               ^
/Users/Bruno/Dropbox/Code/Lean/src/util/thread.h:16:27: note: expanded from
      macro 'LEAN_THREAD_LOCAL'
#define LEAN_THREAD_LOCAL thread_local
                          ^
/Users/Bruno/Dropbox/Code/Lean/src/util/exception.cpp:41:12: error: 
      thread-local storage is unsupported for the current target
    static LEAN_THREAD_LOCAL std::string buffer;
           ^
/Users/Bruno/Dropbox/Code/Lean/src/util/thread.h:16:27: note: expanded from
      macro 'LEAN_THREAD_LOCAL'
#define LEAN_THREAD_LOCAL thread_local
                          ^
2 errors generated.
make[2]: *** [util/CMakeFiles/util.dir/exception.cpp.o] Error 1
make[1]: *** [util/CMakeFiles/util.dir/all] Error 2
make: *** [all] Error 2

add options to setting delay for requesting type information

It should be a configuration option.
The default can be something like 0.4 secs.

add EVAL and SET commands

problem: $(LEAN_VERSION_FILE) on Makefile.common

It does not work on Linux.
I had to remove it to be able to compile the Lean libraries from scratch.

emacs: add supports for SET and EVAL

update tactic framework

tactics such as (apply t) and (exact t) must elaborate t modulo the current goal.

Relaxed identifiers: we should be able to use unicode "letter-like" characters (e.g., beta, delta) in identifiers

update cross-compiler for windows (MXE) to use latest gcc-4.8 snapshot and test thread_local support Edit

Problem

Using x86_64-w64-mingw32-g++ (GCC) 4.8.2 and cross compiling on Linux, Lean is crashing when a thread access thread_local storage.

Reference: http://stackoverflow.com/questions/20277625/why-does-this-usage-of-thread-local-crash?noredirect=1#comment30423002_20277625

Possible Fix

Use the latest snapshot of gcc-4.8 in MXE.

Column numbers are incorrect when we use unicode characters

Create thread for producing type information in --server mode

add "find" command

find gcd -- find theorems that use the constant gcd
find "gcd ?x ?y = gcd ?y ?x"
-- find theorems that match that pattern (anywhere)
find "gcd (gcd _ _) _" -- patterns with anonymous variables
find intro -- find any theorem whose conclusion unifies with your goal
find name: gcd -- find any theorem whose name contains "gcd"

These can be combined, e.g.

find intro gcd
find gcd name: assoc
find gcd "_ + 1"

emacs: redefine lean-mode

• use define-derived-mode and make it based on prog-mode.
• Follow the way idris-mode define its syntax.

feature request: using nested namespaces

Suppose we have

namespace foo
namespace bar
theorem baz
...
end
end

If we import this from another file, we can refer to

foo.bar.baz

If we then say

using foo

it would be convenient if we then could refer to the theorem as

bar.baz

emacs: check buffer-saved when lean-server starts

better way to do "sorry"

add pygments lexer for lean

We can implement a pygments lexer for Lean and make a pull request as Agda and Idris people did:

• Agda: https://bitbucket.org/birkenfeld/pygments-main/pull-request/203/add-literate-agda-lexer/diff
• idris: https://bitbucket.org/birkenfeld/pygments-main/pull-request/210/idris-lexer-added-lexer-for-idris/diff

Github will eventually pick a new version of pygments and use it.

spurious "replacing coercion"

if foo defines a coercion, then every time foo is imported, the "coercion is replaced", e.g.

replacing the coercion from 'num.num' to 'nat.nat'
replacing the coercion from 'num.num' to 'nat.nat'

when the nat library is imported (with files basic, order, sub).

don't create .d file when there is an error

E.g. make produced the following "congr.d":

---

/home/jeremy/projects/lean/library/standard/logic/classes/congr.olean : /home/jeremy/projects/lean/library/standard/logic/classes/cong
r.lean /home/jeremy/projects/lean/library/standard/logic/classes/congr.d error: file 'logic/connectives/function' not found in the LEA
N_PATH

---

Then make produced the following error message:

/home/jeremy/projects/lean/library/standard/logic/classes/congr.d:1: *** multiple target patterns. Stop.

Instead, the first failure should raise the error and not produce the .d file.

lean-mode: flycheck display issue

• remove underline at the tabulate-list
• align error messages to the column

lean-mode: line wrapping in error messages

Currently goals are not pretty-printed, and extend beyond the end of the flycheck error window. (However, when the cursor is on the location, the goal is also displayed at information line at the bottom, which is longer.)

There are three issues here:

(1) goals should be displayed better

(2) the flycheck - lean interface should exchange information about line width

(3) some information (like goals) can appear in different contexts - in the error window, in the information line, and even in a popup - and different formatting is needed for each