NOTE: The CASC ReadMe file is in DOC/Readme
.
This assumes that you have GNU tar, sh and gawk in your search path!
Simple installation of the first-order version (will install executables):
tar -xzf E.tgz
cd E
./configure --bindir=/path/to/EXECDIR
make
make install
/path/to/EXECDIR/eprover -h | more
Simplest installation (in-place):
tar -xzf E.tgz
cd E
./configure
make rebuild
cd PROVER
./eprover -h | more
Read the rest of this file and the fine (if incomplete) manual if
anything fails. There should be a copy of the manual in
DOC/eprover.pdf
.
This version or E supports a fragment of higher-order logic called lambda-free higher-order logic (LFHOL). To build and test the higher-order version, use
./configure --enable-ho
make rebuild
cd PROVER
eprover-ho -h
The recommended command for running E on the file problem.p is
eprover --auto --proof-object problem.p
If you want to try the usually stronger strategy scheduling mode, use
eprover --auto-schedule --proof-object problem.p
Replace eprover
by eprover-ho
for the higher-order-enabled
version.
You can add a time limit of 300 seconds with the option
--cpu-limit=300
and a memory limit of 2 GB with --memory-limit=2048
for all "automatic" modes. You can reduce output with -s
(or
--silent
). As of version 1.9.1, E will try to auto-detect the input
format and adjust the output format accordingly. You can still force
input and output formats via commandline options.
This is the README file for version 2.6 "Floral Guranse" of the E equational theorem prover. This version of E is free software, see the file COPYING for details about the license and the fact that THERE IS NO WARRANTY!
E is an equational theorem prover. That means it is a program that you can stuff a mathematical specification (in many-sorted first-order logic with equality) and a hypothesisconjecture into, and which will then run forever, using up all of your machines resources. Very occasionally it will find a proof for the conjecture and tell you so ;-).
E has been created and is currently maintained by Stephan Schulz,
[email protected], now with the help of several contributors (see
DOC/CONTRIBUTORS
). It is developed and distributed under the GNU
General Public License.
The E homepage can be found at http://www.eprover.org
E can be installed anywhere in the file system, either by a normal user or by the system administrator. By default, the prover will still be compiled as a version that supports first-order logic only.
To install the package, unpack the distribution (if you are reading this, you probably already did):
gunzip -c E.tgz|tar -xvf -
or
(g)tar -xzf E.tgz
(if you have GNU tar)
This should create a directory named E. After unpacking, optionally
edit E/Makefile.vars
to your liking. In particular, if building for
HPUX, comment out the suitable CFLAGS definition (for most systems,
the default definition should be ok). Then change to the E directory:
cd E
Determine if you want to run E from its own build directory or wether you want to install the executables in some other directory EXECDIR. In the first case, run
./configure
otherwise
./configure --bindir=EXECDIR
or, if you also want to install the man-pages into MANDIR,
./configure --bindir=EXECDIR --man-prefix=MANDIR
To enable higher-order-support, add the option --enable-ho
, e.g.
./configure --enable-ho
Then type
make
to compile the libraries and all included programs under the E directory. If you want to install E in a particular EXECDIR, type
make install
You must have write permission in the EXECDIR, so if you install E outside your own home directory, you may need to become root or use sudo.
If you have changed the configuration, you need to rebuild all object and binary files. Run
make rebuild
instead of plain make.
Type
make documentation
to translate the LaTeX documentation (this requires LaTeX2e, pdflatex, and the packages theorem, amssymb and epsfig, which are included in most current LaTeX distributions). The manual should also be included as a pre-compiled PDF.
For some operating systems, especially if you do not have the GNU gcc compiler installed, you may need to edit Makefile.vars manually to select tools and options. If you have any problems, look into E/DOC/PORTING.
If you get into trouble,
make rebuild
will rebuild E completely to your current configuration.
After installation, go to E/PROVER and type
./eprover BOO001-1+rm_eq_rstfp.lop
to see the prover in action. Type
./eprover LUSK6.lop
for a harder example. ./eprover -h
will give you some information and
a list of options.
For impatient people who do not want to read anything:
eprover --auto --memory-limit=<80%_of_your_main_memory> <problem-file>
should give a reasonable performance on a large class of problems (unless your main memory is really small).
The auto mode will perform a heuristic pruning of the axiom set which may result in incompleteness for very large problems (many thousands of axioms). If you need completeness, use
eprover --satauto --memory-limit=<80%_of_your_main_memory> <problem-file>
In general, different proof problems are easy for different strategies. If you run
eprover --auto-schedule --memory-limit=<80%_of_your_main_memory> <problem-file>
or
eprover --satauto-schedule --memory-limit=<80%_of_your_main_memory> <problem-file>
the prover will try a series of strategies on the problem. It assumes
a 300 second run time - if you impose a different one externally, it
is important to let E know via the --cpu-limit=XXX
option so that it
can adjust the schedule.
One of the features of E is the ability to produce semi-readable proofs. To use this, type
eprover --proof-object <other-stuff>
By default, E will now automatically detect the input format (LOP, TPTP-2 or TPTP-3), and will select the matching output format (PCL2 for LOP and TPTP-2 inputs, TPTP-3 for TPTP-3 inputs).
You can check the proof objects in PCL format for correctness using the tool checkproof in the same directory. "checkproof -h" should give you all necessary information. Note that checkproof cannot yet deal with the full first order part, and will skip anything not clausal. Also, support for independent provers in checkproof can be subject to bit-rot, as other systems and interfaces change.
We welcome bug reports and even reasonable questions. If the prover behaves in an unexpected way, please include the following information:
- What did you observe?
- What did you expect?
- The output of
eprover --version
- The full commandline that lead to the unexpected behaviour
- The input file(s) that lead to the unexpected behaviour
Most bug reports should be send to [email protected]. Bug reports with respect to the LFHO-version should be send to [email protected]. Please remember that this is an unpaid volunteer service ;-).