last packaged snapshot. Those are:
--------------------------------------------------
Release Notes for Icarus Verilog Snapshot 20060215
--------------------------------------------------
* Part select of memory words should now work according to
Verilog-2001. This also led to some cleanup of the handling of types
internally, as well as some infrastructure for general arrays.
* Minor fix to parsing of (* *) attributes.
* Fix rounding of reals to integers.
* Clean up some of the vvp engine related to memories. Remove some
dead instructions.
--------------------------------------------------
Release Notes for Icarus Verilog Snapshot 20060409
--------------------------------------------------
the most substantial difference in this snapshot the first signs of
generate support. The compiler now supports generate loops and has
been tested with examples that include wires and gates within the
generate scheme. The regression test suite has very few generate
tests, so any concise self-testing test programs that use generate
would be helpful.
Also, instance arrays that use overridden parameters now work
properly.
Task arguments are a bit more flexible in order to support vendor
(notably Xilinx) models that use more interesting task arguments.
Runtime support for bi-directional ports had some bugs fixed, along
with some other minor run-time bugs. Also, the runtime gains support
for typed parameters. And also, there are some new runtime callbacks
for events and memories.
Parameters had a few types related bugs fixed. They are a bit more
flexible now.
And various minor compilation errors have been fixed. This includes
C/C++ compilation errors fixes, and some configure/Makefile tweaks.
--------------------------------------------------
Release Notes for Icarus Verilog Snapshot 20060618
--------------------------------------------------
Add support for system functions in continuous assignments.
Allow concatenations as arguments to inout ports. This comes with a
small variety of internal part select and concatenation bug fixes.
Fix some bugs in constant propagation through ternary expressions.
Fix broken subtraction if small constants in certain cases.
Fix a few datatype mismatch errors.
Make $readmem give warning when input is inadequate for requested
range.
Fix runtime of nand in continuous assignments.
Fix synchronous user defined primiteves to only follow edges.
Fix a runtime error in some thread delays processing.
Improve limited genvar expression handling.
Start a rework of expression elaboration. Make elaboration aware of
the expression context width when appropriate in order to better
handle expression width and padding.
Fix the make rules for parse.cc to reflect that they come from the
same source. Fix the autoconf.sh to configure the stub target.
Fix portability of the lexor source files on Windows systems. Get rid
of the isatty references.
Make a stub lround when the system version is missing.
--------------------------------------------------
* Release Notes for Snapshot 20060809
--------------------------------------------------
Some handling of real values is improved. Real valued literals are
handled in net contexts (continuous assignment, etc.). Also, modulus
of real operands now works. (This is an extension to the Verilog
standard.)
The power operator (**) now works.
Signed right shift works properly now.
The $sscanf and $fscanf are introduced, and work at least for basic
numeric values.
The release function now works to undo general force statements, and
not just contant force statements.
Delay constants up to 64 bits are supported. This at first doesn't
seem like an issue, but when precisions are mixed, it becomes
surprisingly easy to overflow 32bit delays.
The driver is reworked to pass many preprocessor details through a
temporary file instead of on the command line of a system(3)
call. This prevents confusing and incorrect shell processing of
complex strings passed as values to -D flags.
Various other little fixes.
Changes are:
* Removed all build support for GTK1.
* Dino Ghilardi fixed log.c (the last reminder of GTK1) so
log windows popup don't crash gerbv anymore.
* Dan McMahill added the GCC4 patch.
* Renamed stack_t to macro_stack_t to support darwin/MacOS X
pointed out by Daniel Mack.
* Changed build system to suit new autoconf/automake versions.
Caused gerbv not to find init.scm.
* Fixed a bug so gerbv can handle negative coordinates when using
trailing zeroes.
* Better checking of return values for errors.
* Use setlocale explicit so your locale doesn't screw up sscanf
and strtod to use comma (,) instead of dot (.).
* For distributions there is a new directory called desktop, where
gerbv.desktop and gerbv.png is stored for desktop settings.
Submitted to Debian bug list from Vassilis Pandis for Ubuntu.
since they always need a C compiler, even when the source code is
completely in C++.
For some other packages, stated in the comment that a C compiler is
really not needed.
had actually been ignoring LTCONFIG_OVERRIDE anyway and just using
the default LIBTOOL_OVERRIDE to replace libtool scripts in packages.
This just formalizes the fact that LTCONFIG_OVERRIDE is not used
meaningfully by pkgsrc.
Many improvements since the last packaged version including, but not limited
to:
- enhanced transient timestep control algorithm which is significantly more
accurate than spice now.
- enhanced compatibility with hspice netlists
- named nodes (as opposed to numbered only nodes) work now
- improvements to the bsim3v3 model support
- migration from a custom build system to a GNU autoconf/automake based build
system.
and add a new helper target and script, "show-buildlink3", that outputs
a listing of the buildlink3.mk files included as well as the depth at
which they are included.
For example, "make show-buildlink3" in fonts/Xft2 displays:
zlib
fontconfig
iconv
zlib
freetype2
expat
freetype2
Xrender
renderproto
1.5.4 release was something of a toy, QCad was thoroughly reworked for 2.0
and could now be called usable for many practical purposes.
Versions 2.0.4.0 (initially) to 2.0.5.0 have been usable in wip for a bit
more than a year, and so seem ready to import.
"Yes, that should be fine. Please watch the bulk builds for it and
try to fix any problems that might appear." -wiz@
Updates to pkglint now produce some new suggestions, which I will make and
commit after confirming they don't break anything; at least this version I
already know to build successfully for me.
it will live with other "check" targets run after package installation.
Get rid of SHLIB_HANDLING, whose meaning had mutated over the years
from one thing to another. Currently, it is used to basically note
whether the system's "ldd" command can be usefully run on the package's
binaries and libraries. Rename this variable to CHECK_SHLIBS_SUPPORTED
for more clarity.
CHECK_SHLIBS is now a variable set exclusively by the user in /etc/mk.conf
to note whether the check for missing run-time search paths is performed
after a package is installed. It defaults to "no" unless PKG_DEVELOPER
is set.
