[vmips] VMIPS news for 4/Oct/2002

[Brian R. Gaeke]

Many things have changed in vmips since the last issue of VMIPS news,
so some of these descriptions will be necessarily brief. Feel free to
send mail if you have questions.

VMIPS news for 7/Jun - 4/Oct, 2002
**********************************

Bugs fixed
----------

GDB now gets any errors from accessing memory that doesn't exist, instead
of having them crash VMIPS.

Implement error checking in CPU::debug_fetch_region() and friends.

Fix vmips.cc to use clockspeed option again instead of magic 2000 ns per
instruction value.


Manual & documentation improvements
-----------------------------------

VMIPS manual has been revised in the SPIM keyboard section, to say "byte"
instead of "char".

A "Disconnected operation" subsection has been added to the SPIMconsole
section of the manual.

Added section about using Insight as a front end for VMIPS to the manual.

Added new section to manual about error reporting functions.

A few random documentation updates, including references to other
contributors in AUTHORS and the new web page and e-mail addresses. Make
more references to the bug reporting address.

Generate vmips.pod instead of invoke.pod, so that we get the man page
title right.

Fix some other dependency bugs in doc/Makefile.

doc/makeoptdoc rewritten; now it uses a (slightly more) real parser.

doc/texinfo.tex and doc/texi2pod.pl updated from GNU cvs.

testsuite README updated; now contains slightly fewer buggy recommendations.


Cruft removed
-------------

Remove zschip.cc and zschip.h from the distribution; they never worked
anyway.

Remove cache implementation. We'll start over with a new implementation
next release, if we choose to do so.

Remove periodic.h stuff, serialhost.h stuff.

Clean out some ancient cruft in options.h and sysinclude.h.

Got rid of multiple memory module garbage that was never used.

Got rid of unique_order_deallocate garbage from vmips.h that has not been
used in a long time.

Got rid of old .gdbinit file.


Configuration and Makefiles
---------------------------

Added new configure check for whether __attribute__((noreturn)) works.

Added new configure check for whether __attribute__((format)) works.

Tried to update some of the other configure checks to use more of the
standard autoconf subroutines.

Work around a bug in Automake 1.5 that would complain about assembly source
found but AS not defined ... blah blah.

Renamed some configure checks and cache variables to try to conform to
the standard naming scheme for configure checks as defined in the autoconf
manual.

configure now outputs its output files in sorted order.


Code cleanup
------------

Fixed a zillion typos in comments.

Trade lingering fprintf/abort pairs for fatal_error().

Change fatal_error() and friends to use __attribute__((format)) where
applicable.

Make join.sh a little more portable.

Rewrite Options::usage() so that it looks like a standard GNU application's
--help output.

Rewrite Options::print_package_version() so that it looks like a standard
GNU application's --version output, along with the GNU copyright text.

Move the buffer size #define to remotegdb.h, where it belongs.  Fix some
typos in terminalcontroller.cc.


Testsuite fixes
---------------

Bug in expect scripts for testsuite where expect would hang due to too
many spawn_id's hanging around is fixed.

Bug in expect scripts for testsuite where xfails in outcheck tests would
not be registered is fixed.

Multifile asm testcases now support lack of setup code and
custom-Makefile-fragment test cases are now supported, both using BUILD
RULES statements/comments in .par files.

Errors from trying to run outcheck test cases are now caught by the
framework.

testcase-maker no longer depends on the exact path to expect.

testcase-maker now automatically adds all testcase *.par files to the
distribution.

Testsuite CVS directories are no longer accidentally packaged.


Share and Enjoy,
-Brian


-- 
Brian R. Gaeke, brg@dgate.org -- GnuPG encrypted mail gleefully accepted