use new skolemizer for reconstructing skolemization steps in Isar proofs (because the old skolemizer messes up the order of the Skolem arguments)

rename variable in binder, not just in body

swap Vampire's Skolem arguments to bring them in line with what E and metis's new skolemizer do (helps Isar proof reconstruction in some cases)

tuned comment

NEWS: ML runtime statistics;

merged

more interesting fields;

always enable Future.ML_statistics where this makes sense -- runtime overhead should be negligible;

improved Monitor_Dockable, based on ML_Statistics operations; tuned signature;

get rid of two-year-old hack, now that the "metis" skolemizer no longer gets stuck in HO unification

avoid explosion in higher-order unification algorithm

avoid repeated calls to metis from "resolve_tac" in case of ultimate failure

tuned comment

merged

added standard_frames convenience;

some grouping of standard fields; tuned signature;

some support for chart drawing;

some support for ML statistics content interpretation;

moved files;

moved files;

ML runtime statistics: read properties from build log;

support File.read_gzip as well, in accordance to File.write_gzip; tuned signature;

inline ML statistics into build log;

removed outdated comment;

enable Z3 for full test (cf. 6f48853f08d5);

removed old, unused code

removed whitespace

removed duplicate code

use rpair to avoid swap

generate "obtain" steps corresponding to skolemization inferences

keep E's and Vampire's skolemization steps

tuning

fixed oversensitive Skolem handling (cf. eaa540986291)

added "obtain" to Isar proof construction data structure

tuning

properly take the existential closure of skolems

use "Lazy" to simplify control flow a bit and guarantee single evaluation (at most)

tuning

added 112 to list of known Z3 error codes

actually run Z3 for "SMT_Tests" when "ISABELLE_FULL_TEST" is enabled

added missing certificate file to "ROOT"

more robust report_status: tolerate ML_statistics even if ignored right now, e.g. in batch build;

more liberal edit_control_style: include preceeding control symbol to reduce potential for user surprise;

regenerate certificates -- and use oracle in a few cases where the Z3 proof parser can't deal with Z3 3.2 proofs

Z3's soft timeout is expressed in ms, not in s -- this explains the frequenty "error code 112" failures we had recently

dropped relics of ancient side-entry points

prefer JDialog over JWindow to avoid focus inversion problem on Compiz (e.g. Ubuntu/Unity 12.10): both JDialog and JFrame happen to work, but JFrame does not support parent nesting;

tuned signature;

tuned imports;

include user counters as well;

scala mode is already part of jedit-5.0.0;

tuned;

simplified quick start via "isabelle components -I";

tuned signature -- eliminated obsolete Standard_System;

tuned signature;

recovered generic PIDE build;

updated for release;

updated for release;

updated for release;

misc tuning for release;

recovered Isabelle2012 NEWS from ae12b92c145a, except for e5420161d11d;

updated to smlnj-110.75;

tuned rendering;

ignore markup elements over empty body, which are not well-defined within markup tree and fail to work with merge_disjoint (e.g. multiple inlined positions);

tuned;

more informative error;

tuned -- recovered comments from 791157a4179a;

tuned;

tuned whitespace;

uniform notation for == and \<equiv> (cf. 3e3c2af5e8a5);

tuned;

new theory Library/Finite_Lattice

tuned ML function name

slightly more elegant naming convention (to keep low-level and high-level APIs separated)

tuned ML function names

code checking for Scala is mandatory, since Scala is now required anyway for Isabelle

code checking for Scala is mandatory, since Scala is now required anyway for Isabelle

formally corrected

tuned

uniform parentheses for constructor -- necessary to accomodate scala 10

more explicit name

improved thm order hack, in case the default names are overridden

enable theory learning in MaSh

merged

new version of MaSh

fixed total

new version of MaSh, with theory-level reasoning

tuned

prefer lxbroy10 to evade NFS/hg breakdown seen on various other local machines;

renamed and added lemmas

added simp rule

merged

merged

merged

linearize eval driver, to work around horrible bug in previous implementation

name tuning

merge

better weight functions for MePo/MaSh etc.

merge

more

tuned infix table

crank up default timeout for MaSh ATP learning

removed odd associativity of ==

updated SMT certificates

refined normalization of theorems before giving them to SMT solvers (due to recent changes in rewr_conv, the rewriting of natural-number constants might leave some of them untouched)

tuned "use build timeout": tuples cannot be concatenated

use build timeout (wall clock time, default: 2h)

removed unused usedir_options

removed obsolete setting tweaks: build -s already sets output correctly

plain init_components calls instead of symlinks

tolerate non-existent ISABELLE_OUTPUT

removed obsolete parameter for contrib dir; hard-coding is not a problem

mira: do not hard-code polyml version

discontinued legacy antiquotations and styles

avoid references altogether

catch all parsing errors

no need for tracing

updated MaSh serialization number (to reflect new weights)

synchronize access to shared reference and print proper total

add a timeout in induction rule instantiation