in recent bulk builds.
Release covered-20021214 made. This release is a bug fix release. See list below
for details. Bugs that lead to infinite looping in the score command and segmentation
faults should now be cleared up. Please let me know if there are any other bugs that
need to be addressed before first stable release. Development documentation updated
to match changes in files. Regression suite has been updated quite a bit from last
time. There are now over 125 diagnostics in the regression suite (my goal was to
write about 100 before first stable release).
- Segmentation fault fixes in report command
- Parser can now handle all net types (not just wire). Diagnostics added to regression
suite to verify their proper handling.
- Parser updated to handle net declaration assignments (e.g., wire a = b & c;).
Diagnostics added to verify proper handling.
- Added human-understandable error messages in parser to help identify file and
line number along with a quasi-helpful error message description.
- When parser error is found, Covered exits after parsing phase without continuing
to write CDD file.
- Fixed bug where a multi-bit select expression existed in a module that was
instantiated more than once. Assertion error fired in this case.
- Updated regression suite for VCS testing.
- Fixed bug where parameters were used in modules that were instantiated more than
once.
- Fixed bug that dealt with parameters (see param6.1.v for test case).
- Fixed bug where a delay statement was the last statement in a statement block used
by Covered. Added diagnostics to verify correct behavior.
- Fixed infinite loop problem with db_add_statement function.
- Fixed infinite loop problem with statement_set_stop function.
- Fixed bug with parsing order. When an instance is found for a module that has
already been parsed, the instance was incorrectly being handled. Bug replicated
with instance6.v diagnostic.
- Fixed output of edge-triggered events to add @(...) around the expression (they
were easily confused with other code that could exist on the same line).
- Fixed bug in parser to not allow module to be parsed more than once.
- Fixed bug that lead to an assertion error (see instance6.1.v for test case).
- Fixing bug with calculating list and concatenation lengths when MBIT_SEL
expressions were included.
- Changed Covered's handling of -y directories. Before, all files in these directories
were fed into the parser to look for missing modules. Now, when a module is needed,
the module name is used to find the matching filename in the -y list (basically,
the -y option works like the -y option in Icarus Verilog and VCS). This fix really
streamlined the parsing phase and fixed several bugs.
- Memory declarations are now properly ignored (produced segmentation fault previously).
- Fixed report command to display all lines and expressions in order according to
their line number (the problem is REALLY fixed now).
- Removed hierarchical references from being scored.
All in all, you should notice a huge improvement in the parsing speed, syntax errors are
reported better, more Verilog syntax should be handled properly, the score command will
run a bit faster than before, and the reports should be a bit easier to read. Segmentation
faults and assertion errors should become lesser in number (if not gone altogether?).
I am feeling pretty confident that we are getting close to a stable release as I have
been able to generate a CDD file for a chip that is millions of gates in size (CDD file
was created in the range of 30 - 45 seconds!) Keep the bug reports coming. I have some
things to work on for next release already.
This is the first packaged (in pkgsrc) snapshot after the verilog-0.7
release.
This snapshot adds preliminary support for real variables to the language
to the features already found in verilog-0.7.
tech-pkg@ where the incorrect libtoolize was being invoked. We now pass
in the path to libtoolize via the environment, much like how the other
GNU auto* tools are found in pkgsrc.
This makes it compile with apache-2.0.44.
-make doc installation more robust against reinstalls
-make apache dependency a "full" one -- I don't see what the buildlink
files' default is good for...
bulk-build problem)
Changes: unknown, no information was available in any language or characterset
that I can understand.
Use REPLACE_PERL to obsolete a few patches.
versioned skins, and fall back to ${MASTER_SITE_LOCAL}. A couple go
away, and add a couple new ones, so update ${PKGREVISION}. Closes PR's
pkg/20077 and pkg/20146.
Parts of patch-ag and patch-ah as well as complete patch-aa could be
removed again, they are now included in SA 2.44 (see below).
#### official release announcement ###############
This is a bug-fix release, which fixes the following bugs:
- Backport fix for Bug 1306: Possible buffer overflow in libspamc when
running in BSMTP mode (patch 1.15 -> 1.18)
- Backport workaround from Bug 526: Failed sanity check because of
clobbered STDOUT (patch 1.147 -> 1.148)
- Backport fix for Debian Bug 160206: Insufficient buffer in libspamc
(patch 1.8 -> 1.9)
- Backport fix for warnings in sed_path (patch 1.141 -> 1.142)
- Backport fix for Bug 1127: Existing lowercase x-spam-status header
kills SpamAssassin (patch 1.40 -> 1.41)
- localized %ENV to fix problem where Razor2 erases the PATH so DCC
and
pyzor don't work, etc.
Note that this is *not* 2.50, which offers Bayesian filtering etc. These
bugs are already fixed in the 2.50 CVS tree, but that is not yet ready for
release. This is a stable maintainance release only.
